Michael L. Iuzzolino

Dhruvesh Patel, Hamid Eghbalzadeh, Nitin Kamra et al.

In our pursuit of advancing multi-modal AI assistants capable of guiding users to achieve complex multi-step goals, we propose the task of "Visual Planning for Assistance (VPA)". Given a succinct natural language goal, e.g., "make a shelf", and a video of the user's progress so far, the aim of VPA is to devise a plan, i.e., a sequence of actions such as "sand shelf", "paint shelf", etc. to realize the specified goal. This requires assessing the user's progress from the (untrimmed) video, and relating it to the requirements of natural language goal, i.e., which actions to select and in what order? Consequently, this requires handling long video history and arbitrarily complex action dependencies. To address these challenges, we decompose VPA into video action segmentation and forecasting. Importantly, we experiment by formulating the forecasting step as a multi-modal sequence modeling problem, allowing us to leverage the strength of pre-trained LMs (as the sequence model). This novel approach, which we call Visual Language Model based Planner (VLaMP), outperforms baselines across a suite of metrics that gauge the quality of the generated plans. Furthermore, through comprehensive ablations, we also isolate the value of each component--language pre-training, visual observations, and goal information. We have open-sourced all the data, model checkpoints, and training code.

Reuben Tan, Matthias De Lange, Michael Iuzzolino et al.

Long-term activity forecasting is an especially challenging research problem because it requires understanding the temporal relationships between observed actions, as well as the variability and complexity of human activities. Despite relying on strong supervision via expensive human annotations, state-of-the-art forecasting approaches often generalize poorly to unseen data. To alleviate this issue, we propose Multiscale Video Pretraining (MVP), a novel self-supervised pretraining approach that learns robust representations for forecasting by learning to predict contextualized representations of future video clips over multiple timescales. MVP is based on our observation that actions in videos have a multiscale nature, where atomic actions typically occur at a short timescale and more complex actions may span longer timescales. We compare MVP to state-of-the-art self-supervised video learning approaches on downstream long-term forecasting tasks including long-term action anticipation and video summary prediction. Our comprehensive experiments across the Ego4D and Epic-Kitchens-55/100 datasets demonstrate that MVP out-performs state-of-the-art methods by significant margins. Notably, MVP obtains a relative performance gain of over 20% accuracy in video summary forecasting over existing methods.

Matthias De Lange, Hamid Eghbalzadeh, Reuben Tan et al.

In egocentric action recognition a single population model is typically trained and subsequently embodied on a head-mounted device, such as an augmented reality headset. While this model remains static for new users and environments, we introduce an adaptive paradigm of two phases, where after pretraining a population model, the model adapts on-device and online to the user's experience. This setting is highly challenging due to the change from population to user domain and the distribution shifts in the user's data stream. Coping with the latter in-stream distribution shifts is the focus of continual learning, where progress has been rooted in controlled benchmarks but challenges faced in real-world applications often remain unaddressed. We introduce EgoAdapt, a benchmark for real-world egocentric action recognition that facilitates our two-phased adaptive paradigm, and real-world challenges naturally occur in the egocentric video streams from Ego4d, such as long-tailed action distributions and large-scale classification over 2740 actions. We introduce an evaluation framework that directly exploits the user's data stream with new metrics to measure the adaptation gain over the population model, online generalization, and hindsight performance. In contrast to single-stream evaluation in existing works, our framework proposes a meta-evaluation that aggregates the results from 50 independent user streams. We provide an extensive empirical study for finetuning and experience replay.

Weichao Mao, Ruta Desai, Michael Louis Iuzzolino et al.

Given video demonstrations and paired narrations of an at-home procedural task such as changing a tire, we present an approach to extract the underlying task structure -- relevant actions and their temporal dependencies -- via action-centric task graphs. Learnt structured representations from our method, Action Dynamics Task Graphs (ADTG), can then be used for understanding such tasks in unseen videos of humans performing them. Furthermore, ADTG can enable providing user-centric guidance to humans in these tasks, either for performing them better or for learning new tasks. Specifically, we show how ADTG can be used for: (1) tracking an ongoing task, (2) recommending next actions, and (3) planning a sequence of actions to accomplish a procedural task. We compare against state-of-the-art Neural Task Graph method and demonstrate substantial gains on 18 procedural tasks from the CrossTask dataset, including 30.1% improvement in task tracking accuracy and 20.3% accuracy gain in next action prediction.

Yuxuan Li, Vijay Veerabadran, Michael L. Iuzzolino et al.

We introduce EgoToM, a new video question-answering benchmark that extends Theory-of-Mind (ToM) evaluation to egocentric domains. Using a causal ToM model, we generate multi-choice video QA instances for the Ego4D dataset to benchmark the ability to predict a camera wearer's goals, beliefs, and next actions. We study the performance of both humans and state of the art multimodal large language models (MLLMs) on these three interconnected inference problems. Our evaluation shows that MLLMs achieve close to human-level accuracy on inferring goals from egocentric videos. However, MLLMs (including the largest ones we tested with over 100B parameters) fall short of human performance when inferring the camera wearers' in-the-moment belief states and future actions that are most consistent with the unseen video future. We believe that our results will shape the future design of an important class of egocentric digital assistants which are equipped with a reasonable model of the user's internal mental states.

Xinru Wang, Mengjie Yu, Hannah Nguyen et al.

Large Language Models (LLMs) have shown remarkable potential in recommending everyday actions as personal AI assistants, while Explainable AI (XAI) techniques are being increasingly utilized to help users understand why a recommendation is given. Personal AI assistants today are often located on ultra-small devices such as smartwatches, which have limited screen space. The verbosity of LLM-generated explanations, however, makes it challenging to deliver glanceable LLM explanations on such ultra-small devices. To address this, we explored 1) spatially structuring an LLM's explanation text using defined contextual components during prompting and 2) presenting temporally adaptive explanations to users based on confidence levels. We conducted a user study to understand how these approaches impacted user experiences when interacting with LLM recommendations and explanations on ultra-small devices. The results showed that structured explanations reduced users' time to action and cognitive load when reading an explanation. Always-on structured explanations increased users' acceptance of AI recommendations. However, users were less satisfied with structured explanations compared to unstructured ones due to their lack of sufficient, readable details. Additionally, adaptively presenting structured explanations was less effective at improving user perceptions of the AI compared to the always-on structured explanations. Together with users' interview feedback, the results led to design implications to be mindful of when personalizing the content and timing of LLM explanations that are displayed on ultra-small devices.

Ce Zhang, Yale Song, Ruta Desai et al.

Visual Planning for Assistance (VPA) aims to predict a sequence of user actions required to achieve a specified goal based on a video showing the user's progress. Although recent advances in multimodal large language models (MLLMs) have shown promising results in video understanding, long-horizon visual planning remains a challenging problem. We identify two challenges in training large MLLMs for video-based planning tasks: (1) scarcity of procedural annotations, limiting the model's ability to learn procedural task dynamics effectively, and (2) inefficiency of next-token prediction objective to explicitly capture the structured action space for visual planning when compared to free-form, natural language. To tackle data scarcity, we introduce Auxiliary Task Augmentation. We design and train our model on auxiliary tasks relevant to long-horizon video-based planning (e.g., goal prediction) to augment the model's planning ability. To more explicitly model the structured action space unique to visual planning tasks, we leverage Multi-token Prediction, extending traditional next-token prediction by using multiple heads to predict multiple future tokens during training. Our approach, VideoPlan, achieves state-of-the-art VPA performance on the COIN and CrossTask datasets, surpassing prior methods by 7.3% and 3.4%, respectively, when predicting 3 future actions. We further extend our method to the challenging Ego4D Long-term Action Anticipation task, and show that it is on par with the state-of-the-art approaches despite not using specialized egocentric features. Code will be made available.

Tim Lebailly, Vijay Veerabadran, Satwik Kottur et al.

Generative vision-language models (VLMs) exhibit strong high-level image understanding but lack spatially dense alignment between vision and language modalities, as our findings indicate. Orthogonal to advancements in generative VLMs, another line of research has focused on representation learning for vision-language alignment, targeting zero-shot inference for dense tasks like segmentation. In this work, we bridge these two directions by densely aligning images with synthetic descriptions generated by VLMs. Synthetic captions are inexpensive, scalable, and easy to generate, making them an excellent source of high-level semantic understanding for dense alignment methods. Empirically, our approach outperforms prior work on standard zero-shot open-vocabulary segmentation benchmarks/datasets, while also being more data-efficient.

Mengye Ren, Tyler R. Scott, Michael L. Iuzzolino et al.

Real world learning scenarios involve a nonstationary distribution of classes with sequential dependencies among the samples, in contrast to the standard machine learning formulation of drawing samples independently from a fixed, typically uniform distribution. Furthermore, real world interactions demand learning on-the-fly from few or no class labels. In this work, we propose an unsupervised model that simultaneously performs online visual representation learning and few-shot learning of new categories without relying on any class labels. Our model is a prototype-based memory network with a control component that determines when to form a new class prototype. We formulate it as an online mixture model, where components are created with only a single new example, and assignments do not have to be balanced, which permits an approximation to natural imbalanced distributions from uncurated raw data. Learning includes a contrastive loss that encourages different views of the same image to be assigned to the same prototype. The result is a mechanism that forms categorical representations of objects in nonstationary environments. Experiments show that our method can learn from an online stream of visual input data and its learned representations are significantly better at category recognition compared to state-of-the-art self-supervised learning methods.

Michael L. Iuzzolino, Michael C. Mozer, Samy Bengio

Although deep feedforward neural networks share some characteristics with the primate visual system, a key distinction is their dynamics. Deep nets typically operate in serial stages wherein each layer completes its computation before processing begins in subsequent layers. In contrast, biological systems have cascaded dynamics: information propagates from neurons at all layers in parallel but transmission occurs gradually over time, leading to speed-accuracy trade offs even in feedforward architectures. We explore the consequences of biologically inspired parallel hardware by constructing cascaded ResNets in which each residual block has propagation delays but all blocks update in parallel in a stateful manner. Because information transmitted through skip connections avoids delays, the functional depth of the architecture increases over time, yielding anytime predictions that improve with internal-processing time. We introduce a temporal-difference training loss that achieves a strictly superior speed-accuracy profile over standard losses and enables the cascaded architecture to outperform state-of-the-art anytime-prediction methods. The cascaded architecture has intriguing properties, including: it classifies typical instances more rapidly than atypical instances; it is more robust to both persistent and transient noise than is a conventional ResNet; and its time-varying output trace provides a signal that can be exploited to improve information processing and inference.

Mengye Ren, Michael L. Iuzzolino, Michael C. Mozer et al.

We aim to bridge the gap between typical human and machine-learning environments by extending the standard framework of few-shot learning to an online, continual setting. In this setting, episodes do not have separate training and testing phases, and instead models are evaluated online while learning novel classes. As in the real world, where the presence of spatiotemporal context helps us retrieve learned skills in the past, our online few-shot learning setting also features an underlying context that changes throughout time. Object classes are correlated within a context and inferring the correct context can lead to better performance. Building upon this setting, we propose a new few-shot learning dataset based on large scale indoor imagery that mimics the visual experience of an agent wandering within a world. Furthermore, we convert popular few-shot learning approaches into online versions and we also propose a new contextual prototypical memory model that can make use of spatiotemporal contextual information from the recent past.

Michael L. Iuzzolino, Tetsumichi Umada, Nisar R. Ahmed et al.

We investigate how different active learning (AL) query policies coupled with classification uncertainty visualizations affect analyst trust in automated classification systems. A current standard policy for AL is to query the oracle (e.g., the analyst) to refine labels for datapoints where the classifier has the highest uncertainty. This is an optimal policy for the automation system as it yields maximal information gain. However, model-centric policies neglect the effects of this uncertainty on the human component of the system and the consequent manner in which the human will interact with the system post-training. In this paper, we present an empirical study evaluating how AL query policies and visualizations lending transparency to classification influence trust in automated classification of image data. We found that query policy significantly influences an analyst's trust in an image classification system, and we use these results to propose a set of oracle query policies and visualizations for use during AL training phases that can influence analyst trust in classification.

Hamid Reza Vaezi Joze, Amirreza Shaban, Michael L. Iuzzolino et al.

In late fusion, each modality is processed in a separate unimodal Convolutional Neural Network (CNN) stream and the scores of each modality are fused at the end. Due to its simplicity late fusion is still the predominant approach in many state-of-the-art multimodal applications. In this paper, we present a simple neural network module for leveraging the knowledge from multiple modalities in convolutional neural networks. The propose unit, named Multimodal Transfer Module (MMTM), can be added at different levels of the feature hierarchy, enabling slow modality fusion. Using squeeze and excitation operations, MMTM utilizes the knowledge of multiple modalities to recalibrate the channel-wise features in each CNN stream. Despite other intermediate fusion methods, the proposed module could be used for feature modality fusion in convolution layers with different spatial dimensions. Another advantage of the proposed method is that it could be added among unimodal branches with minimum changes in the their network architectures, allowing each branch to be initialized with existing pretrained weights. Experimental results show that our framework improves the recognition accuracy of well-known multimodal networks. We demonstrate state-of-the-art or competitive performance on four datasets that span the task domains of dynamic hand gesture recognition, speech enhancement, and action recognition with RGB and body joints.

Michael Iuzzolino, Yoram Singer, Michael C. Mozer

In human perception and cognition, a fundamental operation that brains perform is interpretation: constructing coherent neural states from noisy, incomplete, and intrinsically ambiguous evidence. The problem of interpretation is well matched to an early and often overlooked architecture, the attractor network---a recurrent neural net that performs constraint satisfaction, imputation of missing features, and clean up of noisy data via energy minimization dynamics. We revisit attractor nets in light of modern deep learning methods and propose a convolutional bipartite architecture with a novel training loss, activation function, and connectivity constraints. We tackle larger problems than have been previously explored with attractor nets and demonstrate their potential for image completion and super-resolution. We argue that this architecture is better motivated than ever-deeper feedforward models and is a viable alternative to more costly sampling-based generative methods on a range of supervised and unsupervised tasks.

Michael L. Iuzzolino, Michael E. Walker, Daniel Szafir

Robots hold promise in many scenarios involving outdoor use, such as search-and-rescue, wildlife management, and collecting data to improve environment, climate, and weather forecasting. However, autonomous navigation of outdoor trails remains a challenging problem. Recent work has sought to address this issue using deep learning. Although this approach has achieved state-of-the-art results, the deep learning paradigm may be limited due to a reliance on large amounts of annotated training data. Collecting and curating training datasets may not be feasible or practical in many situations, especially as trail conditions may change due to seasonal weather variations, storms, and natural erosion. In this paper, we explore an approach to address this issue through virtual-to-real-world transfer learning using a variety of deep learning models trained to classify the direction of a trail in an image. Our approach utilizes synthetic data gathered from virtual environments for model training, bypassing the need to collect a large amount of real images of the outdoors. We validate our approach in three main ways. First, we demonstrate that our models achieve classification accuracies upwards of 95% on our synthetic data set. Next, we utilize our classification models in the control system of a simulated robot to demonstrate feasibility. Finally, we evaluate our models on real-world trail data and demonstrate the potential of virtual-to-real-world transfer learning.