Alexander Wong

Kevin Zhu, Alexander Wong, John McPhee

Current data analysis for the Canadian Olympic fencing team is primarily done manually by coaches and analysts. Due to the highly repetitive, yet dynamic and subtle movements in fencing, manual data analysis can be inefficient and inaccurate. We propose FenceNet as a novel architecture to automate the classification of fine-grained footwork techniques in fencing. FenceNet takes 2D pose data as input and classifies actions using a skeleton-based action recognition approach that incorporates temporal convolutional networks to capture temporal information. We train and evaluate FenceNet on the Fencing Footwork Dataset (FFD), which contains 10 fencers performing 6 different footwork actions for 10-11 repetitions each (652 total videos). FenceNet achieves 85.4% accuracy under 10-fold cross-validation, where each fencer is left out as the test set. This accuracy is within 1% of the current state-of-the-art method, JLJA (86.3%), which selects and fuses features engineered from skeleton data, depth videos, and inertial measurement units. BiFenceNet, a variant of FenceNet that captures the "bidirectionality" of human movement through two separate networks, achieves 87.6% accuracy, outperforming JLJA. Since neither FenceNet nor BiFenceNet requires data from wearable sensors, unlike JLJA, they could be directly applied to most fencing videos, using 2D pose data as input extracted from off-the-shelf 2D human pose estimators. In comparison to JLJA, our methods are also simpler as they do not require manual feature engineering, selection, or fusion.

Robbie Meyer, Alexander Wong

Model pruning can enable the deployment of neural networks in environments with resource constraints. While pruning may have a small effect on the overall performance of the model, it can exacerbate existing biases into the model such that subsets of samples see significantly degraded performance. In this paper, we introduce the performance weighted loss function, a simple modified cross-entropy loss function that can be used to limit the introduction of biases during pruning. Experiments using the CelebA, Fitzpatrick17k and CIFAR-10 datasets demonstrate that the proposed method is a simple and effective tool that can enable existing pruning methods to be used in fairness sensitive contexts. Code used to produce all experiments contained in this paper can be found at https://github.com/robbiemeyer/pw_loss_pruning.

Blake VanBerlo, Brian Li, Alexander Wong et al.

Self-supervised pretraining has been observed to improve performance in supervised learning tasks in medical imaging. This study investigates the utility of self-supervised pretraining prior to conducting supervised fine-tuning for the downstream task of lung sliding classification in M-mode lung ultrasound images. We propose a novel pairwise relationship that couples M-mode images constructed from the same B-mode image and investigate the utility of data augmentation procedure specific to M-mode lung ultrasound. The results indicate that self-supervised pretraining yields better performance than full supervision, most notably for feature extractors not initialized with ImageNet-pretrained weights. Moreover, we observe that including a vast volume of unlabelled data results in improved performance on external validation datasets, underscoring the value of self-supervision for improving generalizability in automatic ultrasound interpretation. To the authors' best knowledge, this study is the first to characterize the influence of self-supervised pretraining for M-mode ultrasound.

Marjan Shahi, David Clausi, Alexander Wong

In the field of computer vision-driven ice hockey analytics, one of the most challenging and least studied tasks is goalie pose estimation. Unlike general human pose estimation, goalie pose estimation is much more complex as it involves not only the detection of keypoints corresponding to the joints of the goalie concealed under thick padding and mask, but also a large number of non-human keypoints corresponding to the large leg pads and gloves worn, the stick, as well as the hockey net. To tackle this challenge, we introduce GoalieNet, a multi-stage deep neural network for jointly estimating the pose of the goalie, their equipment, and the net. Experimental results using NHL benchmark data demonstrate that the proposed GoalieNet can achieve an average of 84\% accuracy across all keypoints, where 22 out of 29 keypoints are detected with more than 80\% accuracy. This indicates that such a joint pose estimation approach can be a promising research direction.

Maximilian Gilles, Yuhao Chen, Tim Robin Winter et al.

Autonomous bin picking poses significant challenges to vision-driven robotic systems given the complexity of the problem, ranging from various sensor modalities, to highly entangled object layouts, to diverse item properties and gripper types. Existing methods often address the problem from one perspective. Diverse items and complex bin scenes require diverse picking strategies together with advanced reasoning. As such, to build robust and effective machine-learning algorithms for solving this complex task requires significant amounts of comprehensive and high quality data. Collecting such data in real world would be too expensive and time prohibitive and therefore intractable from a scalability perspective. To tackle this big, diverse data problem, we take inspiration from the recent rise in the concept of metaverses, and introduce MetaGraspNet, a large-scale photo-realistic bin picking dataset constructed via physics-based metaverse synthesis. The proposed dataset contains 217k RGBD images across 82 different article types, with full annotations for object detection, amodal perception, keypoint detection, manipulation order and ambidextrous grasp labels for a parallel-jaw and vacuum gripper. We also provide a real dataset consisting of over 2.3k fully annotated high-quality RGBD images, divided into 5 levels of difficulties and an unseen object set to evaluate different object and layout properties. Finally, we conduct extensive experiments showing that our proposed vacuum seal model and synthetic dataset achieves state-of-the-art performance and generalizes to real world use-cases.

Carol Xu, Mahmoud Famouri, Gautam Bathla et al.

Photovoltaic cells are electronic devices that convert light energy to electricity, forming the backbone of solar energy harvesting systems. An essential step in the manufacturing process for photovoltaic cells is visual quality inspection using electroluminescence imaging to identify defects such as cracks, finger interruptions, and broken cells. A big challenge faced by industry in photovoltaic cell visual inspection is the fact that it is currently done manually by human inspectors, which is extremely time consuming, laborious, and prone to human error. While deep learning approaches holds great potential to automating this inspection, the hardware resource-constrained manufacturing scenario makes it challenging for deploying complex deep neural network architectures. In this work, we introduce CellDefectNet, a highly efficient attention condenser network designed via machine-driven design exploration specifically for electroluminesence-based photovoltaic cell defect detection on the edge. We demonstrate the efficacy of CellDefectNet on a benchmark dataset comprising of a diversity of photovoltaic cells captured using electroluminescence imagery, achieving an accuracy of ~86.3% while possessing just 410K parameters (~13$\times$ lower than EfficientNet-B0, respectively) and ~115M FLOPs (~12$\times$ lower than EfficientNet-B0) and ~13$\times$ faster on an ARM Cortex A-72 embedded processor when compared to EfficientNet-B0.

Hossein Aboutalebi, Maya Pavlova, Mohammad Javad Shafiee et al.

Since the World Health Organization declared COVID-19 a pandemic in 2020, the global community has faced ongoing challenges in controlling and mitigating the transmission of the SARS-CoV-2 virus, as well as its evolving subvariants and recombinants. A significant challenge during the pandemic has not only been the accurate detection of positive cases but also the efficient prediction of risks associated with complications and patient survival probabilities. These tasks entail considerable clinical resource allocation and attention.In this study, we introduce COVID-Net Biochem, a versatile and explainable framework for constructing machine learning models. We apply this framework to predict COVID-19 patient survival and the likelihood of developing Acute Kidney Injury during hospitalization, utilizing clinical and biochemical data in a transparent, systematic approach. The proposed approach advances machine learning model design by seamlessly integrating domain expertise with explainability tools, enabling model decisions to be based on key biomarkers. This fosters a more transparent and interpretable decision-making process made by machines specifically for medical applications.

Carol Xu, Mahmoud Famouri, Gautam Bathla et al.

Light guide plates are essential optical components widely used in a diverse range of applications ranging from medical lighting fixtures to back-lit TV displays. In this work, we introduce a fully-integrated, high-throughput, high-performance deep learning-driven workflow for light guide plate surface visual quality inspection (VQI) tailored for real-world manufacturing environments. To enable automated VQI on the edge computing within the fully-integrated VQI system, a highly compact deep anti-aliased attention condenser neural network (which we name LightDefectNet) tailored specifically for light guide plate surface defect detection in resource-constrained scenarios was created via machine-driven design exploration with computational and "best-practices" constraints as well as L_1 paired classification discrepancy loss. Experiments show that LightDetectNet achieves a detection accuracy of ~98.2% on the LGPSDD benchmark while having just 770K parameters (~33X and ~6.9X lower than ResNet-50 and EfficientNet-B0, respectively) and ~93M FLOPs (~88X and ~8.4X lower than ResNet-50 and EfficientNet-B0, respectively) and ~8.8X faster inference speed than EfficientNet-B0 on an embedded ARM processor. As such, the proposed deep learning-driven workflow, integrated with the aforementioned LightDefectNet neural network, is highly suited for high-throughput, high-performance light plate surface VQI within real-world manufacturing environments.

Carol Xu, Mahmoud Famouri, Gautam Bathla et al.

Light guide plates are essential optical components widely used in a diverse range of applications ranging from medical lighting fixtures to back-lit TV displays. An essential step in the manufacturing of light guide plates is the quality inspection of defects such as scratches, bright/dark spots, and impurities. This is mainly done in industry through manual visual inspection for plate pattern irregularities, which is time-consuming and prone to human error and thus act as a significant barrier to high-throughput production. Advances in deep learning-driven computer vision has led to the exploration of automated visual quality inspection of light guide plates to improve inspection consistency, accuracy, and efficiency. However, given the cost constraints in visual inspection scenarios, the widespread adoption of deep learning-driven computer vision methods for inspecting light guide plates has been greatly limited due to high computational requirements. In this study, we explore the utilization of machine-driven design exploration with computational and "best-practices" constraints as well as L$_1$ paired classification discrepancy loss to create LightDefectNet, a highly compact deep anti-aliased attention condenser neural network architecture tailored specifically for light guide plate surface defect detection in resource-constrained scenarios. Experiments show that LightDetectNet achieves a detection accuracy of $\sim$98.2% on the LGPSDD benchmark while having just 770K parameters ($\sim$33$\times$ and $\sim$6.9$\times$ lower than ResNet-50 and EfficientNet-B0, respectively) and $\sim$93M FLOPs ($\sim$88$\times$ and $\sim$8.4$\times$ lower than ResNet-50 and EfficientNet-B0, respectively) and $\sim$8.8$\times$ faster inference speed than EfficientNet-B0 on an embedded ARM processor.

Daniel Zhang, Vikram Voleti, Alexander Wong et al.

Creating high-performance generalizable deep neural networks for phytoplankton monitoring requires utilizing large-scale data coming from diverse global water sources. A major challenge to training such networks lies in data privacy, where data collected at different facilities are often restricted from being transferred to a centralized location. A promising approach to overcome this challenge is federated learning, where training is done at site level on local data, and only the model parameters are exchanged over the network to generate a global model. In this study, we explore the feasibility of leveraging federated learning for privacy-preserving training of deep neural networks for phytoplankton classification. More specifically, we simulate two different federated learning frameworks, federated learning (FL) and mutually exclusive FL (ME-FL), and compare their performance to a traditional centralized learning (CL) framework. Experimental results from this study demonstrate the feasibility and potential of federated learning for phytoplankton monitoring.

E. Zhixuan Zeng, Hayden Gunraj, Sheldon Fernandez et al.

Explainability plays a crucial role in providing a more comprehensive understanding of deep learning models' behaviour. This allows for thorough validation of the model's performance, ensuring that its decisions are based on relevant visual indicators and not biased toward irrelevant patterns existing in training data. However, existing methods provide only instance-level explainability, which requires manual analysis of each sample. Such manual review is time-consuming and prone to human biases. To address this issue, the concept of second-order explainable AI (SOXAI) was recently proposed to extend explainable AI (XAI) from the instance level to the dataset level. SOXAI automates the analysis of the connections between quantitative explanations and dataset biases by identifying prevalent concepts. In this work, we explore the use of this higher-level interpretation of a deep neural network's behaviour to allows us to "explain the explainability" for actionable insights. Specifically, we demonstrate for the first time, via example classification and segmentation cases, that eliminating irrelevant concepts from the training set based on actionable insights from SOXAI can enhance a model's performance.

Blake VanBerlo, Brian Li, Jesse Hoey et al.

In this study, we investigated whether self-supervised pretraining could produce a neural network feature extractor applicable to multiple classification tasks in B-mode lung ultrasound analysis. When fine-tuning on three lung ultrasound tasks, pretrained models resulted in an improvement of the average across-task area under the receiver operating curve (AUC) by 0.032 and 0.061 on local and external test sets respectively. Compact nonlinear classifiers trained on features outputted by a single pretrained model did not improve performance across all tasks; however, they did reduce inference time by 49% compared to serial execution of separate fine-tuned models. When training using 1% of the available labels, pretrained models consistently outperformed fully supervised models, with a maximum observed test AUC increase of 0.396 for the task of view classification. Overall, the results indicate that self-supervised pretraining is useful for producing initial weights for lung ultrasound classifiers.

Stone Yun, Alexander Wong

Deep convolutional neural network (CNN) training via iterative optimization has had incredible success in finding optimal parameters. However, modern CNN architectures often contain millions of parameters. Thus, any given model for a single architecture resides in a massive parameter space. Models with similar loss could have drastically different characteristics such as adversarial robustness, generalizability, and quantization robustness. For deep learning on the edge, quantization robustness is often crucial. Finding a model that is quantization-robust can sometimes require significant efforts. Recent works using Graph Hypernetworks (GHN) have shown remarkable performance predicting high-performant parameters of varying CNN architectures. Inspired by these successes, we wonder if the graph representations of GHN-2 can be leveraged to predict quantization-robust parameters as well, which we call GHN-Q. We conduct the first-ever study exploring the use of graph hypernetworks for predicting parameters of unseen quantized CNN architectures. We focus on a reduced CNN search space and find that GHN-Q can in fact predict quantization-robust parameters for various 8-bit quantized CNNs. Decent quantized accuracies are observed even with 4-bit quantization despite GHN-Q not being trained on it. Quantized finetuning of GHN-Q at lower bitwidths may bring further improvements and is currently being explored.

Chi-en Amy Tai, Jason Li, Sriram Kumar et al.

With the growth in capabilities of generative models, there has been growing interest in using photo-realistic renders of common 3D food items to improve downstream tasks such as food printing, nutrition prediction, or management of food wastage. Despite 3D modelling capabilities being more accessible than ever due to the success of NeRF based view-synthesis, such rendering methods still struggle to correctly capture thin food objects, often generating meshes with significant holes. In this study, we present an optimized strategy for enabling improved rendering of thin 3D food models, and demonstrate qualitative improvements in rendering quality. Our method generates the 3D model mesh via a proposed thin-object-optimized differentiable reconstruction method and tailors the strategy at both the data collection and training stages to better handle thin objects. While simple, we find that this technique can be employed for quick and highly consistent capturing of thin 3D objects.

Chi-en Amy Tai, Oustan Ding, Alexander Wong

In the United States, skin cancer ranks as the most commonly diagnosed cancer, presenting a significant public health issue due to its high rates of occurrence and the risk of serious complications if not caught early. Recent advancements in dataset curation and deep learning have shown promise in quick and accurate detection of skin cancer. However, current open-source datasets have significant class imbalances which impedes the effectiveness of these deep learning models. In healthcare, generative artificial intelligence (AI) models have been employed to create synthetic data, addressing data imbalance in datasets by augmenting underrepresented classes and enhancing the overall quality and performance of machine learning models. In this paper, we build on top of previous work by leveraging new advancements in generative AI, notably Stable Diffusion and DreamBooth. We introduce Cancer-Net SCa-Synth, an open access synthetically generated 2D skin lesion dataset for skin cancer classification. Further analysis on the data effectiveness by comparing the ISIC 2020 test set performance for training with and without these synthetic images for a simple model highlights the benefits of leveraging synthetic data to improve performance. Cancer-Net SCa-Synth is publicly available at https://github.com/catai9/Cancer-Net-SCa-Synth as part of a global open-source initiative for accelerating machine learning for cancer care.

Stone Yun, Alexander Wong

Graph Hypernetworks (GHN) can predict the parameters of varying unseen CNN architectures with surprisingly good accuracy at a fraction of the cost of iterative optimization. Following these successes, preliminary research has explored the use of GHNs to predict quantization-robust parameters for 8-bit and 4-bit quantized CNNs. However, this early work leveraged full-precision float32 training and only quantized for testing. We explore the impact of quantization-aware training and/or other quantization-based training strategies on quantized robustness and performance of GHN predicted parameters for low-precision CNNs. We show that quantization-aware training can significantly improve quantized accuracy for GHN predicted parameters of 4-bit quantized CNNs and even lead to greater-than-random accuracy for 2-bit quantized CNNs. These promising results open the door for future explorations such as investigating the use of GHN predicted parameters as initialization for further quantized training of individual CNNs, further exploration of "extreme bitwidth" quantization, and mixed precision quantization schemes.

E. Zhixuan Zeng, Yuhao Chen, Alexander Wong

Object pose estimation is a critical task in robotics for precise object manipulation. However, current techniques heavily rely on a reference 3D object, limiting their generalizability and making it expensive to expand to new object categories. Direct pose predictions also provide limited information for robotic grasping without referencing the 3D model. Keypoint-based methods offer intrinsic descriptiveness without relying on an exact 3D model, but they may lack consistency and accuracy. To address these challenges, this paper proposes ShapeShift, a superquadric-based framework for object pose estimation that predicts the object's pose relative to a primitive shape which is fitted to the object. The proposed framework offers intrinsic descriptiveness and the ability to generalize to arbitrary geometric shapes beyond the training set.

Brennan Gebotys, Alexander Wong, David A. Clausi

Natural policy gradient methods are popular reinforcement learning methods that improve the stability of policy gradient methods by utilizing second-order approximations to precondition the gradient with the inverse of the Fisher-information matrix. However, to the best of the authors' knowledge, there has not been a study that has investigated the effects of different second-order approximations in a comprehensive and systematic manner. To address this, five different second-order approximations were studied and compared across multiple key metrics including performance, stability, sample efficiency, and computation time. Furthermore, hyperparameters which aren't typically acknowledged in the literature are studied including the effect of different batch sizes and optimizing the critic network with the natural gradient. Experimental results show that on average, improved second-order approximations achieve the best performance and that using properly tuned hyperparameters can lead to large improvements in performance and sample efficiency ranging up to +181%. We also make the code in this study available at https://github.com/gebob19/natural-policy-gradient-reinforcement-learning.

Yuhao Chen, E. Zhixuan Zeng, Maximilian Gilles et al.

There has been increasing interest in smart factories powered by robotics systems to tackle repetitive, laborious tasks. One impactful yet challenging task in robotics-powered smart factory applications is robotic grasping: using robotic arms to grasp objects autonomously in different settings. Robotic grasping requires a variety of computer vision tasks such as object detection, segmentation, grasp prediction, pick planning, etc. While significant progress has been made in leveraging of machine learning for robotic grasping, particularly with deep learning, a big challenge remains in the need for large-scale, high-quality RGBD datasets that cover a wide diversity of scenarios and permutations. To tackle this big, diverse data problem, we are inspired by the recent rise in the concept of metaverse, which has greatly closed the gap between virtual worlds and the physical world. Metaverses allow us to create digital twins of real-world manufacturing scenarios and to virtually create different scenarios from which large volumes of data can be generated for training models. In this paper, we present MetaGraspNet: a large-scale benchmark dataset for vision-driven robotic grasping via physics-based metaverse synthesis. The proposed dataset contains 100,000 images and 25 different object types and is split into 5 difficulties to evaluate object detection and segmentation model performance in different grasping scenarios. We also propose a new layout-weighted performance metric alongside the dataset for evaluating object detection and segmentation performance in a manner that is more appropriate for robotic grasp applications compared to existing general-purpose performance metrics. Our benchmark dataset is available open-source on Kaggle, with the first phase consisting of detailed object detection, segmentation, layout annotations, and a layout-weighted performance metric script.

Alexander Wong, Jack Lu, Adam Dorfman et al.

Pulmonary fibrosis is a devastating chronic lung disease that causes irreparable lung tissue scarring and damage, resulting in progressive loss in lung capacity and has no known cure. A critical step in the treatment and management of pulmonary fibrosis is the assessment of lung function decline, with computed tomography (CT) imaging being a particularly effective method for determining the extent of lung damage caused by pulmonary fibrosis. Motivated by this, we introduce Fibrosis-Net, a deep convolutional neural network design tailored for the prediction of pulmonary fibrosis progression from chest CT images. More specifically, machine-driven design exploration was leveraged to determine a strong architectural design for CT lung analysis, upon which we build a customized network design tailored for predicting forced vital capacity (FVC) based on a patient's CT scan, initial spirometry measurement, and clinical metadata. Finally, we leverage an explainability-driven performance validation strategy to study the decision-making behaviour of Fibrosis-Net as to verify that predictions are based on relevant visual indicators in CT images. Experiments using a patient cohort from the OSIC Pulmonary Fibrosis Progression Challenge showed that the proposed Fibrosis-Net is able to achieve a significantly higher modified Laplace Log Likelihood score than the winning solutions on the challenge. Furthermore, explainability-driven performance validation demonstrated that the proposed Fibrosis-Net exhibits correct decision-making behaviour by leveraging clinically-relevant visual indicators in CT images when making predictions on pulmonary fibrosis progress. While Fibrosis-Net is not yet a production-ready clinical assessment solution, we hope that its release in open source manner will encourage researchers, clinicians, and citizen data scientists alike to leverage and build upon it.

James Ren Hou Lee, Maya Pavlova, Mahmoud Famouri et al.

Skin cancer continues to be the most frequently diagnosed form of cancer in the U.S., with not only significant effects on health and well-being but also significant economic costs associated with treatment. A crucial step to the treatment and management of skin cancer is effective skin cancer detection due to strong prognosis when treated at an early stage, with one of the key screening approaches being dermoscopy examination. Motivated by the advances of deep learning and inspired by the open source initiatives in the research community, in this study we introduce CancerNet-SCa, a suite of deep neural network designs tailored for the detection of skin cancer from dermoscopy images that is open source and available to the general public as part of the Cancer-Net initiative. To the best of the authors' knowledge, CancerNet-SCa comprises of the first machine-designed deep neural network architecture designs tailored specifically for skin cancer detection, one of which possessing a self-attention architecture design with attention condensers. Furthermore, we investigate and audit the behaviour of CancerNet-SCa in a responsible and transparent manner via explainability-driven model auditing. While CancerNet-SCa is not a production-ready screening solution, the hope is that the release of CancerNet-SCa in open source, open access form will encourage researchers, clinicians, and citizen data scientists alike to leverage and build upon them.

William McNally, Kanav Vats, Alexander Wong et al.

Neural architecture search has proven to be highly effective in the design of efficient convolutional neural networks that are better suited for mobile deployment than hand-designed networks. Hypothesizing that neural architecture search holds great potential for human pose estimation, we explore the application of neuroevolution, a form of neural architecture search inspired by biological evolution, in the design of 2D human pose networks for the first time. Additionally, we propose a new weight transfer scheme that enables us to accelerate neuroevolution in a flexible manner. Our method produces network designs that are more efficient and more accurate than state-of-the-art hand-designed networks. In fact, the generated networks process images at higher resolutions using less computation than previous hand-designed networks at lower resolutions, allowing us to push the boundaries of 2D human pose estimation. Our base network designed via neuroevolution, which we refer to as EvoPose2D-S, achieves comparable accuracy to SimpleBaseline while being 50% faster and 12.7x smaller in terms of file size. Our largest network, EvoPose2D-L, achieves new state-of-the-art accuracy on the Microsoft COCO Keypoints benchmark, is 4.3x smaller than its nearest competitor, and has similar inference speed. The code is publicly available at https://github.com/wmcnally/evopose2d.

Nolan Dey, Eric Taylor, Alexander Wong et al.

High quality explanations of neural networks (NNs) should exhibit two key properties. Completeness ensures that they accurately reflect a network's function and interpretability makes them understandable to humans. Many existing methods provide explanations of individual neurons within a network. In this work we provide evidence that for AlexNet pretrained on ImageNet, neuron-based explanation methods sacrifice both completeness and interpretability compared to activation principal components. Neurons are a poor basis for AlexNet embeddings because they don't account for the distributed nature of these representations. By examining two quantitative measures of completeness and conducting a user study to measure interpretability, we show the most important principal components provide more complete and interpretable explanations than the most important neurons. Much of the activation variance may be explained by examining relatively few high-variance PCs, as opposed to studying every neuron. These principal components also strongly affect network function, and are significantly more interpretable than neurons. Our findings suggest that explanation methods for networks like AlexNet should avoid using neurons as a basis for embeddings and instead choose a basis, such as principal components, which accounts for the high dimensional and distributed nature of a network's internal representations. Interactive demo and code available at https://ndey96.github.io/neuron-explanations-sacrifice.

Yuhao Chen, Yifan Wu, Linlin Xu et al.

Building footprint extraction in remote sensing data benefits many important applications, such as urban planning and population estimation. Recently, rapid development of Convolutional Neural Networks (CNNs) and open-sourced high resolution satellite building image datasets have pushed the performance boundary further for automated building extractions. However, CNN approaches often generate imprecise building morphologies including noisy edges and round corners. In this paper, we leverage the performance of CNNs, and propose a module that uses prior knowledge of building corners to create angular and concise building polygons from CNN segmentation outputs. We describe a new transform, Relative Gradient Angle Transform (RGA Transform) that converts object contours from time vs. space to time vs. angle. We propose a new shape descriptor, Boundary Orientation Relation Set (BORS), to describe angle relationship between edges in RGA domain, such as orthogonality and parallelism. Finally, we develop an energy minimization framework that makes use of the angle relationship in BORS to straighten edges and reconstruct sharp corners, and the resulting corners create a polygon. Experimental results demonstrate that our method refines CNN output from a rounded approximation to a more clear-cut angular shape of the building footprint.

Aaryam Sharma, Chris Czarnecki, Yuhao Chen et al.

Monitoring dietary intake is a crucial aspect of promoting healthy living. In recent years, advances in computer vision technology have facilitated dietary intake monitoring through the use of images and depth cameras. However, the current state-of-the-art image-based food portion estimation algorithms assume that users take images of their meals one or two times, which can be inconvenient and fail to capture food items that are not visible from a top-down perspective, such as ingredients submerged in a stew. To address these limitations, we introduce an innovative solution that utilizes stationary user-facing cameras to track food items on utensils, not requiring any change of camera perspective after installation. The shallow depth of utensils provides a more favorable angle for capturing food items, and tracking them on the utensil's surface offers a significantly more accurate estimation of dietary intake without the need for post-meal image capture. The system is reliable for estimation of nutritional content of liquid-solid heterogeneous mixtures such as soups and stews. Through a series of experiments, we demonstrate the exceptional potential of our method as a non-invasive, user-friendly, and highly accurate dietary intake monitoring tool.

Dayou Mao, Yuhao Chen, Yifan Wu et al.

One of the main motivations of MTL is to develop neural networks capable of inferring multiple tasks simultaneously. While countless methods have been proposed in the past decade investigating robust model architectures and efficient training algorithms, there is still lack of understanding of these methods when applied on smaller feature extraction backbones, the generalizability of the commonly used fast approximation technique of replacing parameter-level gradients with feature level gradients, and lack of comprehensive understanding of MTL challenges and how one can efficiently and effectively identify the challenges. In this paper, we focus on the aforementioned efficiency aspects of existing MTL methods. We first carry out large-scale experiments of the methods with smaller backbones and on a the MetaGraspNet dataset as a new test ground. We also compare the existing methods with and without using the fast gradient surrogate and empirically study the generalizability of this technique. Lastly, we propose Feature Disentanglement measure as a novel and efficient identifier of the challenges in MTL, and propose Ranking Similarity score as an evaluation metric for different identifiers to prove the faithfulness of our method.

Bavesh Balaji, Jerrin Bright, Sirisha Rambhatla et al.

Unique player identification is a fundamental module in vision-driven sports analytics. Identifying players from broadcast videos can aid with various downstream tasks such as player assessment, in-game analysis, and broadcast production. However, automatic detection of jersey numbers using deep features is challenging primarily due to: a) motion blur, b) low resolution video feed, and c) occlusions. With their recent success in various vision tasks, masked autoencoders (MAEs) have emerged as a superior alternative to conventional feature extractors. However, most MAEs simply zero-out image patches either randomly or focus on where to mask rather than how to mask. Motivated by human vision, we devise a novel domain-guided masking policy for MAEs termed d-MAE to facilitate robust feature extraction in the presence of motion blur for player identification. We further introduce a new spatio-temporal network leveraging our novel d-MAE for unique player identification. We conduct experiments on three large-scale sports datasets, including a curated baseball dataset, the SoccerNet dataset, and an in-house ice hockey dataset. We preprocess the datasets using an upgraded keyframe identification (KfID) module by focusing on frames containing jersey numbers. Additionally, we propose a keyframe-fusion technique to augment keyframes, preserving spatial and temporal context. Our spatio-temporal network showcases significant improvements, surpassing the current state-of-the-art by 8.58%, 4.29%, and 1.20% in the test set accuracies, respectively. Rigorous ablations highlight the effectiveness of our domain-guided masking approach and the refined KfID module, resulting in performance enhancements of 1.48% and 1.84% respectively, compared to original architectures.

Blake VanBerlo, Alexander Wong, Jesse Hoey et al.

Self-supervised learning (SSL) is one strategy for addressing the paucity of labelled data in medical imaging by learning representations from unlabelled images. Contrastive and non-contrastive SSL methods produce learned representations that are similar for pairs of related images. Such pairs are commonly constructed by randomly distorting the same image twice. The videographic nature of ultrasound offers flexibility for defining the similarity relationship between pairs of images. In this study, we investigated the effect of utilizing proximal, distinct images from the same B-mode ultrasound video as pairs for SSL. Additionally, we introduced a sample weighting scheme that increases the weight of closer image pairs and demonstrated how it can be integrated into SSL objectives. Named Intra-Video Positive Pairs (IVPP), the method surpassed previous ultrasound-specific contrastive learning methods' average test accuracy on COVID-19 classification with the POCUS dataset by $\ge 1.3\%$. Detailed investigations of IVPP's hyperparameters revealed that some combinations of IVPP hyperparameters can lead to improved or worsened performance, depending on the downstream task. Guidelines for practitioners were synthesized based on the results, such as the merit of IVPP with task-specific hyperparameters, and the improved performance of contrastive methods for ultrasound compared to non-contrastive counterparts.

Joshua Li, Fernando Jose Pena Cantu, Emily Yu et al.

Video Scene Graph Generation (VidSGG) is an important topic in understanding dynamic kitchen environments. Current models for VidSGG require extensive training to produce scene graphs. Recently, Vision Language Models (VLM) and Vision Foundation Models (VFM) have demonstrated impressive zero-shot capabilities in a variety of tasks. However, VLMs like Gemini struggle with the dynamics for VidSGG, failing to maintain stable object identities across frames. To overcome this limitation, we propose SAMJAM, a zero-shot pipeline that combines SAM2's temporal tracking with Gemini's semantic understanding. SAM2 also improves upon Gemini's object grounding by producing more accurate bounding boxes. In our method, we first prompt Gemini to generate a frame-level scene graph. Then, we employ a matching algorithm to map each object in the scene graph with a SAM2-generated or SAM2-propagated mask, producing a temporally-consistent scene graph in dynamic environments. Finally, we repeat this process again in each of the following frames. We empirically demonstrate that SAMJAM outperforms Gemini by 8.33% in mean recall on the EPIC-KITCHENS and EPIC-KITCHENS-100 datasets.

Chi-en Amy Tai, Alexander Wong

In 2020, prostate cancer saw a staggering 1.4 million new cases, resulting in over 375,000 deaths. The accurate identification of clinically significant prostate cancer is crucial for delivering effective treatment to patients. Consequently, there has been a surge in research exploring the application of deep neural networks to predict clinical significance based on magnetic resonance images. However, these networks demand extensive datasets to attain optimal performance. Recently, transfer learning emerged as a technique that leverages acquired features from a domain with richer data to enhance the performance of a domain with limited data. In this paper, we investigate the improvement of clinically significant prostate cancer prediction in T2-weighted images through transfer learning from breast cancer. The results demonstrate a remarkable improvement of over 30% in leave-one-out cross-validation accuracy.

Chi-en Amy Tai, Alexander Wong

Peptide de novo sequencing is a method used to reconstruct amino acid sequences from tandem mass spectrometry data without relying on existing protein sequence databases. Traditional deep learning approaches, such as Casanovo, mainly utilize autoregressive decoders and predict amino acids sequentially. Subsequently, they encounter cascading errors and fail to leverage high-confidence regions effectively. To address these issues, this paper investigates using diffusion decoders adapted for the discrete data domain. These decoders provide a different approach, allowing sequence generation to start from any peptide segment, thereby enhancing prediction accuracy. We experiment with three different diffusion decoder designs, knapsack beam search, and various loss functions. We find knapsack beam search did not improve performance metrics and simply replacing the transformer decoder with a diffusion decoder lowered performance. Although peptide precision and recall were still 0, the best diffusion decoder design with the DINOISER loss function obtained a statistically significant improvement in amino acid recall by 0.373 compared to the baseline autoregressive decoder-based Casanovo model. These findings highlight the potential of diffusion decoders to not only enhance model sensitivity but also drive significant advancements in peptide de novo sequencing.

E. Zhixuan Zeng, Yuhao Chen, Alexander Wong

Image generation techniques, particularly latent diffusion models, have exploded in popularity in recent years. Many techniques have been developed to manipulate and clarify the semantic concepts these large-scale models learn, offering crucial insights into biases and concept relationships. However, these techniques are often only validated in conventional realms of human or animal faces and artistic style transitions. The food domain offers unique challenges through complex compositions and regional biases, which can shed light on the limitations and opportunities within existing methods. Through the lens of food imagery, we analyze both qualitative and quantitative patterns within a concept traversal technique. We reveal measurable insights into the model's ability to capture and represent the nuances of culinary diversity, while also identifying areas where the model's biases and limitations emerge.

Parthipan Siva, Alexander Wong, Patricia Hewston et al.

With an aging population, numerous assistive and monitoring technologies are under development to enable older adults to age in place. To facilitate aging in place predicting risk factors such as falls, and hospitalization and providing early interventions are important. Much of the work on ambient monitoring for risk prediction has centered on gait speed analysis, utilizing privacy-preserving sensors like radar. Despite compelling evidence that monitoring step length, in addition to gait speed, is crucial for predicting risk, radar-based methods have not explored step length measurement in the home. Furthermore, laboratory experiments on step length measurement using radars are limited to proof of concept studies with few healthy subjects. To address this gap, a radar-based step length measurement system for the home is proposed based on detection and tracking using radar point cloud, followed by Doppler speed profiling of the torso to obtain step lengths in the home. The proposed method was evaluated in a clinical environment, involving 35 frail older adults, to establish its validity. Additionally, the method was assessed in people's homes, with 21 frail older adults who had participated in the clinical assessment. The proposed radar-based step length measurement method was compared to the gold standard Zeno Walkway Gait Analysis System, revealing a 4.5cm/8.3% error in a clinical setting. Furthermore, it exhibited excellent reliability (ICC(2,k)=0.91, 95% CI 0.82 to 0.96) in uncontrolled home settings. The method also proved accurate in uncontrolled home settings, as indicated by a strong agreement (ICC(3,k)=0.81 (95% CI 0.53 to 0.92)) between home measurements and in-clinic assessments.

Saeejith Nair, Mohammad Javad Shafiee, Alexander Wong

We present DARLEI, a framework that combines evolutionary algorithms with parallelized reinforcement learning for efficiently training and evolving populations of UNIMAL agents. Our approach utilizes Proximal Policy Optimization (PPO) for individual agent learning and pairs it with a tournament selection-based generational learning mechanism to foster morphological evolution. By building on Nvidia's Isaac Gym, DARLEI leverages GPU accelerated simulation to achieve over 20x speedup using just a single workstation, compared to previous work which required large distributed CPU clusters. We systematically characterize DARLEI's performance under various conditions, revealing factors impacting diversity of evolved morphologies. For example, by enabling inter-agent collisions within the simulator, we find that we can simulate some multi-agent interactions between the same morphology, and see how it influences individual agent capabilities and long-term evolutionary adaptation. While current results demonstrate limited diversity across generations, we hope to extend DARLEI in future work to include interactions between diverse morphologies in richer environments, and create a platform that allows for coevolving populations and investigating emergent behaviours in them. Our source code is also made publicly at https://saeejithnair.github.io/darlei.

Mohammad Javad Shafiee, Mahmoud Famouri, Gautam Bathla et al.

A critical aspect in the manufacturing process is the visual quality inspection of manufactured components for defects and flaws. Human-only visual inspection can be very time-consuming and laborious, and is a significant bottleneck especially for high-throughput manufacturing scenarios. Given significant advances in the field of deep learning, automated visual quality inspection can lead to highly efficient and reliable detection of defects and flaws during the manufacturing process. However, deep learning-driven visual inspection methods often necessitate significant computational resources, thus limiting throughput and act as a bottleneck to widespread adoption for enabling smart factories. In this study, we investigated the utilization of a machine-driven design exploration approach to create TinyDefectNet, a highly compact deep convolutional network architecture tailored for high-throughput manufacturing visual quality inspection. TinyDefectNet comprises of just ~427K parameters and has a computational complexity of ~97M FLOPs, yet achieving a detection accuracy of a state-of-the-art architecture for the task of surface defect detection on the NEU defect benchmark dataset. As such, TinyDefectNet can achieve the same level of detection performance at 52$\times$ lower architectural complexity and 11x lower computational complexity. Furthermore, TinyDefectNet was deployed on an AMD EPYC 7R32, and achieved 7.6x faster throughput using the native Tensorflow environment and 9x faster throughput using AMD ZenDNN accelerator library. Finally, explainability-driven performance validation strategy was conducted to ensure correct decision-making behaviour was exhibited by TinyDefectNet to improve trust in its usage by operators and inspectors.

Brennan Gebotys, Alexander Wong, David A. Clausi

Advancements in attention mechanisms have led to significant performance improvements in a variety of areas in machine learning due to its ability to enable the dynamic modeling of temporal sequences. A particular area in computer vision that is likely to benefit greatly from the incorporation of attention mechanisms in video action recognition. However, much of the current research's focus on attention mechanisms have been on spatial and temporal attention, which are unable to take advantage of the inherent motion found in videos. Motivated by this, we develop a new attention mechanism called Motion Aware Attention (M2A) that explicitly incorporates motion characteristics. More specifically, M2A extracts motion information between consecutive frames and utilizes attention to focus on the motion patterns found across frames to accurately recognize actions in videos. The proposed M2A mechanism is simple to implement and can be easily incorporated into any neural network backbone architecture. We show that incorporating motion mechanisms with attention mechanisms using the proposed M2A mechanism can lead to a +15% to +26% improvement in top-1 accuracy across different backbone architectures, with only a small increase in computational complexity. We further compared the performance of M2A with other state-of-the-art motion and attention mechanisms on the Something-Something V1 video action recognition benchmark. Experimental results showed that M2A can lead to further improvements when combined with other temporal mechanisms and that it outperforms other motion-only or attention-only mechanisms by as much as +60% in top-1 accuracy for specific classes in the benchmark.

Audrey Chung, Mahmoud Famouri, Andrew Hryniowski et al.

The COVID-19 pandemic continues to have a devastating global impact, and has placed a tremendous burden on struggling healthcare systems around the world. Given the limited resources, accurate patient triaging and care planning is critical in the fight against COVID-19, and one crucial task within care planning is determining if a patient should be admitted to a hospital's intensive care unit (ICU). Motivated by the need for transparent and trustworthy ICU admission clinical decision support, we introduce COVID-Net Clinical ICU, a neural network for ICU admission prediction based on patient clinical data. Driven by a transparent, trust-centric methodology, the proposed COVID-Net Clinical ICU was built using a clinical dataset from Hospital Sirio-Libanes comprising of 1,925 COVID-19 patient records, and is able to predict when a COVID-19 positive patient would require ICU admission with an accuracy of 96.9% to facilitate better care planning for hospitals amidst the on-going pandemic. We conducted system-level insight discovery using a quantitative explainability strategy to study the decision-making impact of different clinical features and gain actionable insights for enhancing predictive performance. We further leveraged a suite of trust quantification metrics to gain deeper insights into the trustworthiness of COVID-Net Clinical ICU. By digging deeper into when and why clinical predictive models makes certain decisions, we can uncover key factors in decision making for critical clinical decision support tasks such as ICU admission prediction and identify the situations under which clinical predictive models can be trusted for greater accountability.

Maziar Gomrokchi, Susan Amin, Hossein Aboutalebi et al.

While significant research advances have been made in the field of deep reinforcement learning, there have been no concrete adversarial attack strategies in literature tailored for studying the vulnerability of deep reinforcement learning algorithms to membership inference attacks. In such attacking systems, the adversary targets the set of collected input data on which the deep reinforcement learning algorithm has been trained. To address this gap, we propose an adversarial attack framework designed for testing the vulnerability of a state-of-the-art deep reinforcement learning algorithm to a membership inference attack. In particular, we design a series of experiments to investigate the impact of temporal correlation, which naturally exists in reinforcement learning training data, on the probability of information leakage. Moreover, we compare the performance of \emph{collective} and \emph{individual} membership attacks against the deep reinforcement learning algorithm. Experimental results show that the proposed adversarial attack framework is surprisingly effective at inferring data with an accuracy exceeding $84\%$ in individual and $97\%$ in collective modes in three different continuous control Mujoco tasks, which raises serious privacy concerns in this regard. Finally, we show that the learning state of the reinforcement learning algorithm influences the level of privacy breaches significantly.

George Michalopoulos, Ian McKillop, Alexander Wong et al.

Lexical substitution is the task of generating meaningful substitutes for a word in a given textual context. Contextual word embedding models have achieved state-of-the-art results in the lexical substitution task by relying on contextual information extracted from the replaced word within the sentence. However, such models do not take into account structured knowledge that exists in external lexical databases. We introduce LexSubCon, an end-to-end lexical substitution framework based on contextual embedding models that can identify highly accurate substitute candidates. This is achieved by combining contextual information with knowledge from structured lexical resources. Our approach involves: (i) introducing a novel mix-up embedding strategy in the creation of the input embedding of the target word through linearly interpolating the pair of the target input embedding and the average embedding of its probable synonyms; (ii) considering the similarity of the sentence-definition embeddings of the target word and its proposed candidates; and, (iii) calculating the effect of each substitution in the semantics of the sentence through a fine-tuned sentence similarity model. Our experiments show that LexSubCon outperforms previous state-of-the-art methods on LS07 and CoInCo benchmark datasets that are widely used for lexical substitution tasks.

Hossein Aboutalebi, Mohammad Javad Shafiee, Michelle Karg et al.

Despite the significant advances in deep learning over the past decade, a major challenge that limits the wide-spread adoption of deep learning has been their fragility to adversarial attacks. This sensitivity to making erroneous predictions in the presence of adversarially perturbed data makes deep neural networks difficult to adopt for certain real-world, mission-critical applications. While much of the research focus has revolved around adversarial example creation and adversarial hardening, the area of performance measures for assessing adversarial robustness is not well explored. Motivated by this, this study presents the concept of residual error, a new performance measure for not only assessing the adversarial robustness of a deep neural network at the individual sample level, but also can be used to differentiate between adversarial and non-adversarial examples to facilitate for adversarial example detection. Furthermore, we introduce a hybrid model for approximating the residual error in a tractable manner. Experimental results using the case of image classification demonstrates the effectiveness and efficacy of the proposed residual error metric for assessing several well-known deep neural network architectures. These results thus illustrate that the proposed measure could be a useful tool for not only assessing the robustness of deep neural networks used in mission-critical scenarios, but also in the design of adversarially robust models.

William McNally, Pascale Walters, Kanav Vats et al.

Existing multi-camera solutions for automatic scorekeeping in steel-tip darts are very expensive and thus inaccessible to most players. Motivated to develop a more accessible low-cost solution, we present a new approach to keypoint detection and apply it to predict dart scores from a single image taken from any camera angle. This problem involves detecting multiple keypoints that may be of the same class and positioned in close proximity to one another. The widely adopted framework for regressing keypoints using heatmaps is not well-suited for this task. To address this issue, we instead propose to model keypoints as objects. We develop a deep convolutional neural network around this idea and use it to predict dart locations and dartboard calibration points within an overall pipeline for automatic dart scoring, which we call DeepDarts. Additionally, we propose several task-specific data augmentation strategies to improve the generalization of our method. As a proof of concept, two datasets comprising 16k images originating from two different dartboard setups were manually collected and annotated to evaluate the system. In the primary dataset containing 15k images captured from a face-on view of the dartboard using a smartphone, DeepDarts predicted the total score correctly in 94.7% of the test images. In a second more challenging dataset containing limited training data (830 images) and various camera angles, we utilize transfer learning and extensive data augmentation to achieve a test accuracy of 84.0%. Because DeepDarts relies only on single images, it has the potential to be deployed on edge devices, giving anyone with a smartphone access to an automatic dart scoring system for steel-tip darts. The code and datasets are available.

Mehrnaz Fani, Pascale Berunelle Walters, David A. Clausi et al.

In this work, an automatic and simple framework for hockey ice-rink localization from broadcast videos is introduced. First, video is broken into video-shots by a hierarchical partitioning of the video frames, and thresholding based on their histograms. To localize the frames on the ice-rink model, a ResNet18-based regressor is implemented and trained, which regresses to four control points on the model in a frame-by-frame fashion. This leads to the projection jittering problem in the video. To overcome this, in the inference phase, the trajectory of the control points on the ice-rink model are smoothed, for all the consecutive frames of a given video-shot, by convolving a Hann window with the achieved coordinates. Finally, the smoothed homography matrix is computed by using the direct linear transform on the four pairs of corresponding points. A hockey dataset for training and testing the regressor is gathered. The results show success of this simple and comprehensive procedure for localizing the hockey ice-rink and addressing the problem of jittering without affecting the accuracy of homography estimation.

Saad Abbasi, Mahmoud Famouri, Mohammad Javad Shafiee et al.

Human operators often diagnose industrial machinery via anomalous sounds. Automated acoustic anomaly detection can lead to reliable maintenance of machinery. However, deep learning-driven anomaly detection methods often require an extensive amount of computational resources which prohibits their deployment in factories. Here we explore a machine-driven design exploration strategy to create OutlierNets, a family of highly compact deep convolutional autoencoder network architectures featuring as few as 686 parameters, model sizes as small as 2.7 KB, and as low as 2.8 million FLOPs, with a detection accuracy matching or exceeding published architectures with as many as 4 million parameters. Furthermore, CPU-accelerated latency experiments show that the OutlierNet architectures can achieve as much as 21x lower latency than published networks.

Ahmadreza Jeddi, Mohammad Javad Shafiee, Alexander Wong

Adversarial Training (AT) with Projected Gradient Descent (PGD) is an effective approach for improving the robustness of the deep neural networks. However, PGD AT has been shown to suffer from two main limitations: i) high computational cost, and ii) extreme overfitting during training that leads to reduction in model generalization. While the effect of factors such as model capacity and scale of training data on adversarial robustness have been extensively studied, little attention has been paid to the effect of a very important parameter in every network optimization on adversarial robustness: the learning rate. In particular, we hypothesize that effective learning rate scheduling during adversarial training can significantly reduce the overfitting issue, to a degree where one does not even need to adversarially train a model from scratch but can instead simply adversarially fine-tune a pre-trained model. Motivated by this hypothesis, we propose a simple yet very effective adversarial fine-tuning approach based on a $\textit{slow start, fast decay}$ learning rate scheduling strategy which not only significantly decreases computational cost required, but also greatly improves the accuracy and robustness of a deep neural network. Experimental results show that the proposed adversarial fine-tuning approach outperforms the state-of-the-art methods on CIFAR-10, CIFAR-100 and ImageNet datasets in both test accuracy and the robustness, while reducing the computational cost by 8-10$\times$. Furthermore, a very important benefit of the proposed adversarial fine-tuning approach is that it enables the ability to improve the robustness of any pre-trained deep neural network without needing to train the model from scratch, which to the best of the authors' knowledge has not been previously demonstrated in research literature.

Andrew Hryniowski, Alexander Wong

The quantitative analysis of information structure through a deep neural network (DNN) can unveil new insights into the theoretical performance of DNN architectures. Two very promising avenues of research towards quantitative information structure analysis are: 1) layer similarity (LS) strategies focused on the inter-layer feature similarity, and 2) intrinsic dimensionality (ID) strategies focused on layer-wise data dimensionality using pairwise information. Inspired by both LS and ID strategies for quantitative information structure analysis, we introduce two novel complimentary methods for inter-layer information similarity assessment premised on the interesting idea of studying a data sample's neighbourhood dynamics as it traverses through a DNN. More specifically, we introduce the concept of Nearest Neighbour Topological Similarity (NNTS) for quantifying the information topology similarity between layers of a DNN. Furthermore, we introduce the concept of Nearest Neighbour Topological Persistence (NNTP) for quantifying the inter-layer persistence of data neighbourhood relationships throughout a DNN. The proposed strategies facilitate the efficient inter-layer information similarity assessment by leveraging only local topological information, and we demonstrate their efficacy in this study by performing analysis on a deep convolutional neural network architecture on image data to study the insights that can be gained with respect to the theoretical performance of a DNN.

Alexander Wong, Andrew Hryniowski, Xiao Yu Wang

The success of deep learning in recent years have led to a significant increase in interest and prevalence for its adoption to tackle financial services tasks. One particular question that often arises as a barrier to adopting deep learning for financial services is whether the developed financial deep learning models are fair in their predictions, particularly in light of strong governance and regulatory compliance requirements in the financial services industry. A fundamental aspect of fairness that has not been explored in financial deep learning is the concept of trust, whose variations may point to an egocentric view of fairness and thus provide insights into the fairness of models. In this study we explore the feasibility and utility of a multi-scale trust quantification strategy to gain insights into the fairness of a financial deep learning model, particularly under different scenarios at different scales. More specifically, we conduct multi-scale trust quantification on a deep neural network for the purpose of credit card default prediction to study: 1) the overall trustworthiness of the model 2) the trust level under all possible prediction-truth relationships, 3) the trust level across the spectrum of possible predictions, 4) the trust level across different demographic groups (e.g., age, gender, and education), and 5) distribution of overall trust for an individual prediction scenario. The insights for this proof-of-concept study demonstrate that such a multi-scale trust quantification strategy may be helpful for data scientists and regulators in financial services as part of the verification and certification of financial deep learning solutions to gain insights into fairness and trust of these solutions.

James Ren Hou Lee, Alexander Wong

The ability to interpret social cues comes naturally for most people, but for those living with Autism Spectrum Disorder (ASD), some experience a deficiency in this area. This paper presents the development of a multimodal augmented reality (AR) system which combines the use of computer vision and deep convolutional neural networks (CNN) in order to assist individuals with the detection and interpretation of facial expressions in social settings. The proposed system, which we call AEGIS (Augmented-reality Expression Guided Interpretation System), is an assistive technology deployable on a variety of user devices including tablets, smartphones, video conference systems, or smartglasses, showcasing its extreme flexibility and wide range of use cases, to allow integration into daily life with ease. Given a streaming video camera source, each real-world frame is passed into AEGIS, processed for facial bounding boxes, and then fed into our novel deep convolutional time windowed neural network (TimeConvNet). We leverage both spatial and temporal information in order to provide an accurate expression prediction, which is then converted into its corresponding visualization and drawn on top of the original video frame. The system runs in real-time, requires minimal set up and is simple to use. With the use of AEGIS, we can assist individuals living with ASD to learn to better identify expressions and thus improve their social experiences.

George Michalopoulos, Yuanxin Wang, Hussam Kaka et al.

Contextual word embedding models, such as BioBERT and Bio_ClinicalBERT, have achieved state-of-the-art results in biomedical natural language processing tasks by focusing their pre-training process on domain-specific corpora. However, such models do not take into consideration expert domain knowledge. In this work, we introduced UmlsBERT, a contextual embedding model that integrates domain knowledge during the pre-training process via a novel knowledge augmentation strategy. More specifically, the augmentation on UmlsBERT with the Unified Medical Language System (UMLS) Metathesaurus was performed in two ways: i) connecting words that have the same underlying `concept' in UMLS, and ii) leveraging semantic group knowledge in UMLS to create clinically meaningful input embeddings. By applying these two strategies, UmlsBERT can encode clinical domain knowledge into word embeddings and outperform existing domain-specific models on common named-entity recognition (NER) and clinical natural language inference clinical NLP tasks.

Andrew Hryniowski, Xiao Yu Wang, Alexander Wong

The advances and successes in deep learning in recent years have led to considerable efforts and investments into its widespread ubiquitous adoption for a wide variety of applications, ranging from personal assistants and intelligent navigation to search and product recommendation in e-commerce. With this tremendous rise in deep learning adoption comes questions about the trustworthiness of the deep neural networks that power these applications. Motivated to answer such questions, there has been a very recent interest in trust quantification. In this work, we introduce the concept of trust matrix, a novel trust quantification strategy that leverages the recently introduced question-answer trust metric by Wong et al. to provide deeper, more detailed insights into where trust breaks down for a given deep neural network given a set of questions. More specifically, a trust matrix defines the expected question-answer trust for a given actor-oracle answer scenario, allowing one to quickly spot areas of low trust that needs to be addressed to improve the trustworthiness of a deep neural network. The proposed trust matrix is simple to calculate, humanly interpretable, and to the best of the authors' knowledge is the first to study trust at the actor-oracle answer level. We further extend the concept of trust densities with the notion of conditional trust densities. We experimentally leverage trust matrices to study several well-known deep neural network architectures for image recognition, and further study the trust density and conditional trust densities for an interesting actor-oracle answer scenario. The results illustrate that trust matrices, along with conditional trust densities, can be useful tools in addition to the existing suite of trust quantification metrics for guiding practitioners and regulators in creating and certifying deep learning solutions for trusted operation.

Alexander Wong, Mahmoud Famouri, Mohammad Javad Shafiee

While significant advances in deep learning has resulted in state-of-the-art performance across a large number of complex visual perception tasks, the widespread deployment of deep neural networks for TinyML applications involving on-device, low-power image recognition remains a big challenge given the complexity of deep neural networks. In this study, we introduce AttendNets, low-precision, highly compact deep neural networks tailored for on-device image recognition. More specifically, AttendNets possess deep self-attention architectures based on visual attention condensers, which extends on the recently introduced stand-alone attention condensers to improve spatial-channel selective attention. Furthermore, AttendNets have unique machine-designed macroarchitecture and microarchitecture designs achieved via a machine-driven design exploration strategy. Experimental results on ImageNet$_{50}$ benchmark dataset for the task of on-device image recognition showed that AttendNets have significantly lower architectural and computational complexity when compared to several deep neural networks in research literature designed for efficiency while achieving highest accuracies (with the smallest AttendNet achieving $\sim$7.2% higher accuracy, while requiring $\sim$3$\times$ fewer multiply-add operations, $\sim$4.17$\times$ fewer parameters, and $\sim$16.7$\times$ lower weight memory requirements than MobileNet-V1). Based on these promising results, AttendNets illustrate the effectiveness of visual attention condensers as building blocks for enabling various on-device visual perception tasks for TinyML applications.