Sudeep Sarkar

Shenyuan Liang, Mauricio Pamplona Segundo, Sathyanarayanan N. Aakur et al.

This paper focuses on the statistical analysis of shapes of data objects called shape graphs, a set of nodes connected by articulated curves with arbitrary shapes. A critical need here is a constrained registration of points (nodes to nodes, edges to edges) across objects. This, in turn, requires optimization over the permutation group, made challenging by differences in nodes (in terms of numbers, locations) and edges (in terms of shapes, placements, and sizes) across objects. This paper tackles this registration problem using a novel neural-network architecture and involves an unsupervised loss function developed using the elastic shape metric for curves. This architecture results in (1) state-of-the-art matching performance and (2) an order of magnitude reduction in the computational cost relative to baseline approaches. We demonstrate the effectiveness of the proposed approach using both simulated data and real-world 2D and 3D shape graphs. Code and data will be made publicly available after review to foster research.

Cole Hill, Mauricio Pamplona Segundo, Sudeep Sarkar

Deep learning research has made many biometric recognition solution viable, but it requires vast training data to achieve real-world generalization. Unlike other biometric traits, such as face and ear, gait samples cannot be easily crawled from the web to form massive unconstrained datasets. As the human body has been extensively studied for different digital applications, one can rely on prior shape knowledge to overcome data scarcity. This work follows the recent trend of fitting a 3D deformable body model into gait videos using deep neural networks to obtain disentangled shape and pose representations for each frame. To enforce temporal consistency in the network, we introduce a new Linear Dynamical Systems (LDS) module and loss based on Koopman operator theory, which provides an unsupervised motion regularization for the periodic nature of gait, as well as a predictive capacity for extending gait sequences. We compare LDS to the traditional adversarial training approach and use the USF HumanID and CASIA-B datasets to show that LDS can obtain better accuracy with less training data. Finally, we also show that our 3D modeling approach is much better than other 3D gait approaches in overcoming viewpoint variation under normal, bag-carrying and clothing change conditions.

Sujal Vijayaraghavan, Redwan Alqasemi, Rajiv Dubey et al.

Robot grasp typically follows five stages: object detection, object localisation, object pose estimation, grasp pose estimation, and grasp planning. We focus on object pose estimation. Our approach relies on three pieces of information: multiple views of the object, the camera's extrinsic parameters at those viewpoints, and 3D CAD models of objects. The first step involves a standard deep learning backbone (FCN ResNet) to estimate the object label, semantic segmentation, and a coarse estimate of the object pose with respect to the camera. Our novelty is using a refinement module that starts from the coarse pose estimate and refines it by optimisation through differentiable rendering. This is a purely vision-based approach that avoids the need for other information such as point cloud or depth images. We evaluate our object pose estimation approach on the ShapeNet dataset and show improvements over the state of the art. We also show that the estimated object pose results in 99.65% grasp accuracy with the ground truth grasp candidates on the Object Clutter Indoor Dataset (OCID) Grasp dataset, as computed using standard practice.

Mauricio Pamplona Segundo, Allan Pinto, Rodrigo Minetto et al.

This work introduces a novel solution to measure economic activity through remote sensing for a wide range of spatial areas. We hypothesized that disturbances in human behavior caused by major life-changing events leave signatures in satellite imagery that allows devising relevant image-based indicators to estimate their impacts and support decision-makers. We present a case study for the COVID-19 coronavirus outbreak, which imposed severe mobility restrictions and caused worldwide disruptions, using flying airplane detection around the 30 busiest airports in Europe to quantify and analyze the lockdown's effects and post-lockdown recovery. Our solution won the Rapid Action Coronavirus Earth observation (RACE) upscaling challenge, sponsored by the European Space Agency and the European Commission, and now integrates the RACE dashboard. This platform combines satellite data and artificial intelligence to promote a progressive and safe reopening of essential activities. Code and CNN models are available at https://github.com/maups/covid19-custom-script-contest

Ramy Mounir, Roman Gula, Jörn Theuerkauf et al.

Using offline training schemes, researchers have tackled the event segmentation problem by providing full or weak-supervision through manually annotated labels or self-supervised epoch-based training. Most works consider videos that are at most 10's of minutes long. We present a self-supervised perceptual prediction framework capable of temporal event segmentation by building stable representations of objects over time and demonstrate it on long videos, spanning several days. The approach is deceptively simple but quite effective. We rely on predictions of high-level features computed by a standard deep learning backbone. For prediction, we use an LSTM, augmented with an attention mechanism, trained in a self-supervised manner using the prediction error. The self-learned attention maps effectively localize and track the event-related objects in each frame. The proposed approach does not require labels. It requires only a single pass through the video, with no separate training set. Given the lack of datasets of very long videos, we demonstrate our method on video from 10 days (254 hours) of continuous wildlife monitoring data that we had collected with required permissions. We find that the approach is robust to various environmental conditions such as day/night conditions, rain, sharp shadows, and windy conditions. For the task of temporally locating events, we had an 80% recall rate at 20% false-positive rate for frame-level segmentation. At the activity level, we had an 80% activity recall rate for one false activity detection every 50 minutes. We will make the dataset, which is the first of its kind, and the code available to the research community.

Rodrigo Minetto, Mauricio Pamplona Segundo, Gilbert Rotich et al.

The COVID-19 outbreak forced governments worldwide to impose lockdowns and quarantines to prevent virus transmission. As a consequence, there are disruptions in human and economic activities all over the globe. The recovery process is also expected to be rough. Economic activities impact social behaviors, which leave signatures in satellite images that can be automatically detected and classified. Satellite imagery can support the decision-making of analysts and policymakers by providing a different kind of visibility into the unfolding economic changes. In this work, we use a deep learning approach that combines strategic location sampling and an ensemble of lightweight convolutional neural networks (CNNs) to recognize specific elements in satellite images that could be used to compute economic indicators based on it, automatically. This CNN ensemble framework ranked third place in the US Department of Defense xView challenge, the most advanced benchmark for object detection in satellite images. We show the potential of our framework for temporal analysis using the US IARPA Function Map of the World (fMoW) dataset. We also show results on real examples of different sites before and after the COVID-19 outbreak to illustrate different measurable indicators. Our code and annotated high-resolution aerial scenes before and after the outbreak are available on GitHub (https://github.com/maups/covid19-satellite-analysis).

Rodrigo Minetto, Mauricio Pamplona Segundo, Sudeep Sarkar

We describe in this paper Hydra, an ensemble of convolutional neural networks (CNN) for geospatial land classification. The idea behind Hydra is to create an initial CNN that is coarsely optimized but provides a good starting pointing for further optimization, which will serve as the Hydra's body. Then, the obtained weights are fine-tuned multiple times with different augmentation techniques, crop styles, and classes weights to form an ensemble of CNNs that represent the Hydra's heads. By doing so, we prompt convergence to different endpoints, which is a desirable aspect for ensembles. With this framework, we were able to reduce the training time while maintaining the classification performance of the ensemble. We created ensembles for our experiments using two state-of-the-art CNN architectures, ResNet and DenseNet. We have demonstrated the application of our Hydra framework in two datasets, FMOW and NWPU-RESISC45, achieving results comparable to the state-of-the-art for the former and the best reported performance so far for the latter. Code and CNN models are available at https://github.com/maups/hydra-fmow

Sathyanarayanan N. Aakur, Sudeep Sarkar

Event perception tasks such as recognizing and localizing actions in streaming videos are essential for scaling to real-world application contexts. We tackle the problem of learning actor-centered representations through the notion of continual hierarchical predictive learning to localize actions in streaming videos without the need for training labels and outlines for the objects in the video. We propose a framework driven by the notion of hierarchical predictive learning to construct actor-centered features by attention-based contextualization. The key idea is that predictable features or objects do not attract attention and hence do not contribute to the action of interest. Experiments on three benchmark datasets show that the approach can learn robust representations for localizing actions using only one epoch of training, i.e., a single pass through the streaming video. We show that the proposed approach outperforms unsupervised and weakly supervised baselines while offering competitive performance to fully supervised approaches. Additionally, we extend the model to multi-actor settings to recognize group activities while localizing the multiple, plausible actors. We also show that it generalizes to out-of-domain data with limited performance degradation.

Sathyanarayanan N. Aakur, Sudeep Sarkar

The problem of action recognition involves locating the action in the video, both over time and spatially in the image. The dominant current approaches use supervised learning to solve this problem, and require large amounts of annotated training data, in the form of frame-level bounding box annotations around the region of interest. In this paper, we present a new approach based on continual learning that uses feature-level predictions for self-supervision. It does not require any training annotations in terms of frame-level bounding boxes. The approach is inspired by cognitive models of visual event perception that propose a prediction-based approach to event understanding. We use a stack of LSTMs coupled with CNN encoder, along with novel attention mechanisms, to model the events in the video and use this model to predict high-level features for the future frames. The prediction errors are used to continuously learn the parameters of the models. This self-supervised framework is not complicated as other approaches but is very effective in learning robust visual representations for both labeling and localization. It should be noted that the approach outputs in a streaming fashion, requiring only a single pass through the video, making it amenable for real-time processing. We demonstrate this on three datasets - UCF Sports, JHMDB, and THUMOS'13 and show that the proposed approach outperforms weakly-supervised and unsupervised baselines and obtains competitive performance compared to fully supervised baselines. Finally, we show that the proposed framework can generalize to egocentric videos and obtain state-of-the-art results in unsupervised gaze prediction.

Xiaoyang Guo, Anuj Srivastava, Sudeep Sarkar

Complex analyses involving multiple, dependent random quantities often lead to graphical models - a set of nodes denoting variables of interest, and corresponding edges denoting statistical interactions between nodes. To develop statistical analyses for graphical data, especially towards generative modeling, one needs mathematical representations and metrics for matching and comparing graphs, and subsequent tools, such as geodesics, means, and covariances. This paper utilizes a quotient structure to develop efficient algorithms for computing these quantities, leading to useful statistical tools, including principal component analysis, statistical testing, and modeling. We demonstrate the efficacy of this framework using datasets taken from several problem areas, including letters, biochemical structures, and social networks.

Sathyanarayanan N. Aakur, Sudeep Sarkar

Question answering has seen significant advances in recent times, especially with the introduction of increasingly bigger transformer-based models pre-trained on massive amounts of data. While achieving impressive results on many benchmarks, their performances appear to be proportional to the amount of training data available in the target domain. In this work, we explore the ability of current question-answering models to generalize - to both other domains as well as with restricted training data. We find that large amounts of training data are necessary, both for pre-training as well as fine-tuning to a task, for the models to perform well on the designated task. We introduce a novel abductive reasoning approach based on Grenander's Pattern Theory framework to provide self-supervised domain adaptation cues or "pseudo-labels," which can be used instead of expensive human annotations. The proposed self-supervised training regimen allows for effective domain adaptation without losing performance compared to fully supervised baselines. Extensive experiments on two publicly available benchmarks show the efficacy of the proposed approach. We show that neural networks models trained using self-labeled data can retain up to $75\%$ of the performance of models trained on large amounts of human-annotated training data.

Žiga Emeršič, Aruna Kumar S. V., B. S. Harish et al.

This paper presents a summary of the 2019 Unconstrained Ear Recognition Challenge (UERC), the second in a series of group benchmarking efforts centered around the problem of person recognition from ear images captured in uncontrolled settings. The goal of the challenge is to assess the performance of existing ear recognition techniques on a challenging large-scale ear dataset and to analyze performance of the technology from various viewpoints, such as generalization abilities to unseen data characteristics, sensitivity to rotations, occlusions and image resolution and performance bias on sub-groups of subjects, selected based on demographic criteria, i.e. gender and ethnicity. Research groups from 12 institutions entered the competition and submitted a total of 13 recognition approaches ranging from descriptor-based methods to deep-learning models. The majority of submissions focused on ensemble based methods combining either representations from multiple deep models or hand-crafted with learned image descriptors. Our analysis shows that methods incorporating deep learning models clearly outperform techniques relying solely on hand-crafted descriptors, even though both groups of techniques exhibit similar behaviour when it comes to robustness to various covariates, such presence of occlusions, changes in (head) pose, or variability in image resolution. The results of the challenge also show that there has been considerable progress since the first UERC in 2017, but that there is still ample room for further research in this area.

Sathyanarayanan N. Aakur, Sudeep Sarkar

Temporal segmentation of long videos is an important problem, that has largely been tackled through supervised learning, often requiring large amounts of annotated training data. In this paper, we tackle the problem of self-supervised temporal segmentation of long videos that alleviate the need for any supervision. We introduce a self-supervised, predictive learning framework that draws inspiration from cognitive psychology to segment long, visually complex videos into individual, stable segments that share the same semantics. We also introduce a new adaptive learning paradigm that helps reduce the effect of catastrophic forgetting in recurrent neural networks. Extensive experiments on three publicly available datasets - Breakfast Actions, 50 Salads, and INRIA Instructional Videos datasets show the efficacy of the proposed approach. We show that the proposed approach is able to outperform weakly-supervised and other unsupervised learning approaches by up to 24% and have competitive performance compared to fully supervised approaches. We also show that the proposed approach is able to learn highly discriminative features that help improve action recognition when used in a representation learning paradigm.

Gilbert Rotich, Rodrigo Minetto, Sudeep Sarkar

We describe a strategy for detection and classification of man-made objects in large high-resolution satellite photos under computational resource constraints. We detect and classify candidate objects by using five pipelines of convolutional neural network processing (CNN), run in parallel. Each pipeline has its own unique strategy for fine tunning parameters, proposal region filtering, and dealing with image scales. The conflicting region proposals are merged based on region confidence and not just based on overlap areas, which improves the quality of the final bounding-box regions selected. We demonstrate this strategy using the recent xView challenge, which is a complex benchmark with more than 1,100 high-resolution images, spanning 800,000 aerial objects around the world covering a total area of 1,400 square kilometers at 0.3 meter ground sample distance. To tackle the resource-constrained problem posed by the xView challenge, where inferences are restricted to be on CPU with 8GB memory limit, we used lightweight CNN's trained with the single shot detector algorithm. Our approach was competitive on sequestered sets; it was ranked third.

Ghada Zamzmi, Gabriel Ruiz, Matthew Shreve et al.

We address the problem of suppressing facial expressions in videos because expressions can hinder the retrieval of important information in applications such as face recognition. To achieve this, we present an optical strain suppression method that removes any facial expression without requiring training for a specific expression. For each frame in a video, an optical strain map that provides the strain magnitude value at each pixel is generated; this strain map is then utilized to neutralize the expression by replacing pixels of high strain values with pixels from a reference face frame. Experimental results of testing the method on various expressions namely happiness, sadness, and anger for two publicly available data sets (i.e., BU-4DFE and AM-FED) show the ability of our method in suppressing facial expressions.

Earnest E. Hansley, Mauricio Pamplona Segundo, Sudeep Sarkar

We present an unconstrained ear recognition framework that outperforms state-of-the-art systems in different publicly available image databases. To this end, we developed CNN-based solutions for ear normalization and description, we used well-known handcrafted descriptors, and we fused learned and handcrafted features to improve recognition. We designed a two-stage landmark detector that successfully worked under untrained scenarios. We used the results generated to perform a geometric image normalization that boosted the performance of all evaluated descriptors. Our CNN descriptor outperformed other CNN-based works in the literature, specially in more difficult scenarios. The fusion of learned and handcrafted matchers appears to be complementary as it achieved the best performance in all experiments. The obtained results outperformed all other reported results for the UERC challenge, which contains the most difficult database nowadays.

Sathyanarayanan N. Aakur, Fillipe DM de Souza, Sudeep Sarkar

A deeper understanding of video activities extends beyond recognition of underlying concepts such as actions and objects: constructing deep semantic representations requires reasoning about the semantic relationships among these concepts, often beyond what is directly observed in the data. To this end, we propose an energy minimization framework that leverages large-scale commonsense knowledge bases, such as ConceptNet, to provide contextual cues to establish semantic relationships among entities directly hypothesized from video signal. We mathematically express this using the language of Grenander's canonical pattern generator theory. We show that the use of prior encoded commonsense knowledge alleviate the need for large annotated training datasets and help tackle imbalance in training through prior knowledge. Using three different publicly available datasets - Charades, Microsoft Visual Description Corpus and Breakfast Actions datasets, we show that the proposed model can generate video interpretations whose quality is better than those reported by state-of-the-art approaches, which have substantial training needs. Through extensive experiments, we show that the use of commonsense knowledge from ConceptNet allows the proposed approach to handle various challenges such as training data imbalance, weak features, and complex semantic relationships and visual scenes.

Mona Fathollahi Ghezelghieh, Rangachar Kasturi, Sudeep Sarkar

The objective of this work is to estimate 3D human pose from a single RGB image. Extracting image representations which incorporate both spatial relation of body parts and their relative depth plays an essential role in accurate3D pose reconstruction. In this paper, for the first time, we show that camera viewpoint in combination to 2D joint lo-cations significantly improves 3D pose accuracy without the explicit use of perspective geometry mathematical models.To this end, we train a deep Convolutional Neural Net-work (CNN) to learn categorical camera viewpoint. To make the network robust against clothing and body shape of the subject in the image, we utilized 3D computer rendering to synthesize additional training images. We test our framework on the largest 3D pose estimation bench-mark, Human3.6m, and achieve up to 20% error reduction compared to the state-of-the-art approaches that do not use body part segmentation.