Ross Goroshin

Jesse Farebrother, Joshua Greaves, Rishabh Agarwal et al. · deepmind

Auxiliary tasks improve the representations learned by deep reinforcement learning agents. Analytically, their effect is reasonably well understood; in practice, however, their primary use remains in support of a main learning objective, rather than as a method for learning representations. This is perhaps surprising given that many auxiliary tasks are defined procedurally, and hence can be treated as an essentially infinite source of information about the environment. Based on this observation, we study the effectiveness of auxiliary tasks for learning rich representations, focusing on the setting where the number of tasks and the size of the agent's network are simultaneously increased. For this purpose, we derive a new family of auxiliary tasks based on the successor measure. These tasks are easy to implement and have appealing theoretical properties. Combined with a suitable off-policy learning rule, the result is a representation learning algorithm that can be understood as extending Mahadevan & Maggioni (2007)'s proto-value functions to deep reinforcement learning -- accordingly, we call the resulting object proto-value networks. Through a series of experiments on the Arcade Learning Environment, we demonstrate that proto-value networks produce rich features that may be used to obtain performance comparable to established algorithms, using only linear approximation and a small number (~4M) of interactions with the environment's reward function.

Mahan Fathi, Jonathan Pilault, Orhan Firat et al. · mila

State space models (SSMs) have shown impressive results on tasks that require modeling long-range dependencies and efficiently scale to long sequences owing to their subquadratic runtime complexity. Originally designed for continuous signals, SSMs have shown superior performance on a plethora of tasks, in vision and audio; however, SSMs still lag Transformer performance in Language Modeling tasks. In this work, we propose a hybrid layer named Block-State Transformer (BST), that internally combines an SSM sublayer for long-range contextualization, and a Block Transformer sublayer for short-term representation of sequences. We study three different, and completely parallelizable, variants that integrate SSMs and block-wise attention. We show that our model outperforms similar Transformer-based architectures on language modeling perplexity and generalizes to longer sequences. In addition, the Block-State Transformer demonstrates more than tenfold increase in speed at the layer level compared to the Block-Recurrent Transformer when model parallelization is employed.

Mahan Fathi, Clement Gehring, Jonathan Pilault et al. · mila

Koopman representations aim to learn features of nonlinear dynamical systems (NLDS) which lead to linear dynamics in the latent space. Theoretically, such features can be used to simplify many problems in modeling and control of NLDS. In this work we study autoencoder formulations of this problem, and different ways they can be used to model dynamics, specifically for future state prediction over long horizons. We discover several limitations of predicting future states in the latent space and propose an inference-time mechanism, which we refer to as Periodic Reencoding, for faithfully capturing long term dynamics. We justify this method both analytically and empirically via experiments in low and high dimensional NLDS.

Sebastian Jung, Artem Zholus, Martin Sundermeyer et al.

Tracking-Any-Point (TAP) models aim to track any point through a video which is a crucial task in AR/XR and robotics applications. The recently introduced TAPNext approach proposes an end-to-end, recurrent transformer architecture to track points frame-by-frame in a purely online fashion -- demonstrating competitive performance at minimal latency. However, we show that TAPNext struggles with longer video sequences and also frequently fails to re-detect query points that reappear after being occluded or leaving the frame. In this work, we present TAPNext++, a model that tracks points in sequences that are orders of magnitude longer while preserving the low memory and compute footprint of the architecture. We train the recurrent video transformer using several data-driven solutions, including training on long 1024-frame sequences enabled by sequence parallelism techniques. We highlight that re-detection performance is a blind spot in the current literature and introduce a new metric, Re-Detection Average Jaccard ($AJ_{RD}$), to explicitly evaluate tracking on re-appearing points. To improve re-detection of points, we introduce tailored geometric augmentations, such as periodic roll that simulates point re-entries, and supervising occluded points. We demonstrate that recurrent transformers can be substantially improved for point tracking and set a new state-of-the-art on multiple benchmarks. Model and code can be found at https://tap-next-plus-plus.github.io.

Ross Goroshin, Alex Wilson, Andrew Lamb et al.

Project Sunroof estimates the solar potential of residential buildings using high quality aerial data. That is, it estimates the potential solar energy (and associated financial savings) that can be captured by buildings if solar panels were to be installed on their roofs. Unfortunately its coverage is limited by the lack of high resolution digital surface map (DSM) data. We present a deep learning approach that bridges this gap by enhancing widely available low-resolution data, thereby dramatically increasing the coverage of Sunroof. We also present some ongoing efforts to potentially improve accuracy even further by replacing certain algorithmic components of the Sunroof processing pipeline with deep learning.

João Carreira, Dilara Gokay, Michael King et al.

Scaling has not yet been convincingly demonstrated for pure self-supervised learning from video. However, prior work has focused evaluations on semantic-related tasks $\unicode{x2013}$ action classification, ImageNet classification, etc. In this paper we focus on evaluating self-supervised learning on non-semantic vision tasks that are more spatial (3D) and temporal (+1D = 4D), such as camera pose estimation, point and object tracking, and depth estimation. We show that by learning from very large video datasets, masked auto-encoding (MAE) with transformer video models actually scales, consistently improving performance on these 4D tasks, as model size increases from 20M all the way to the largest by far reported self-supervised video model $\unicode{x2013}$ 22B parameters. Rigorous apples-to-apples comparison with many recent image and video models demonstrates the benefits of scaling 4D representations. Pretrained models are available at https://github.com/google-deepmind/representations4d .

Vishal Batchu, Alex Wilson, Betty Peng et al.

The transition to renewable energy, particularly solar, is key to mitigating climate change. Google's Solar API aids this transition by estimating solar potential from aerial imagery, but its impact is constrained by geographical coverage. This paper proposes expanding the API's reach using satellite imagery, enabling global solar potential assessment. We tackle challenges involved in building a Digital Surface Model (DSM) and roof instance segmentation from lower resolution and single oblique views using deep learning models. Our models, trained on aligned satellite and aerial datasets, produce 25cm DSMs and roof segments. With ~1m DSM MAE on buildings, ~5deg roof pitch error and ~56% IOU on roof segmentation, they significantly enhance the Solar API's potential to promote solar adoption.

Viorica Pătrăucean, Xu Owen He, Joseph Heyward et al. · deepmind

We propose a novel block for video modelling. It relies on a time-space-channel factorisation with dedicated blocks for each dimension: gated linear recurrent units (LRUs) perform information mixing over time, self-attention layers perform mixing over space, and MLPs over channels. The resulting architecture TRecViT performs well on sparse and dense tasks, trained in supervised or self-supervised regimes. Notably, our model is causal and outperforms or is on par with a pure attention model ViViT-L on large scale video datasets (SSv2, Kinetics400), while having $3\times$ less parameters, $12\times$ smaller memory footprint, and $5\times$ lower FLOPs count. Code and checkpoints will be made available online at https://github.com/google-deepmind/trecvit.

Cristina Vasconcelos, Hugo Larochelle, Vincent Dumoulin et al.

We investigate the impact of aliasing on generalization in Deep Convolutional Networks and show that data augmentation schemes alone are unable to prevent it due to structural limitations in widely used architectures. Drawing insights from frequency analysis theory, we take a closer look at ResNet and EfficientNet architectures and review the trade-off between aliasing and information loss in each of their major components. We show how to mitigate aliasing by inserting non-trainable low-pass filters at key locations, particularly where networks lack the capacity to learn them. These simple architectural changes lead to substantial improvements in generalization on i.i.d. and even more on out-of-distribution conditions, such as image classification under natural corruptions on ImageNet-C [11] and few-shot learning on Meta-Dataset [26]. State-of-the art results are achieved on both datasets without introducing additional trainable parameters and using the default hyper-parameters of open source codebases.

Cristina Vasconcelos, Hugo Larochelle, Vincent Dumoulin et al.

Image pre-processing in the frequency domain has traditionally played a vital role in computer vision and was even part of the standard pipeline in the early days of deep learning. However, with the advent of large datasets, many practitioners concluded that this was unnecessary due to the belief that these priors can be learned from the data itself. Frequency aliasing is a phenomenon that may occur when sub-sampling any signal, such as an image or feature map, causing distortion in the sub-sampled output. We show that we can mitigate this effect by placing non-trainable blur filters and using smooth activation functions at key locations, particularly where networks lack the capacity to learn them. These simple architectural changes lead to substantial improvements in out-of-distribution generalization on both image classification under natural corruptions on ImageNet-C [10] and few-shot learning on Meta-Dataset [17], without introducing additional trainable parameters and using the default hyper-parameters of open source codebases.

Carl Doersch, Pauline Luc, Yi Yang et al.

To endow models with greater understanding of physics and motion, it is useful to enable them to perceive how solid surfaces move and deform in real scenes. This can be formalized as Tracking-Any-Point (TAP), which requires the algorithm to track any point on solid surfaces in a video, potentially densely in space and time. Large-scale groundtruth training data for TAP is only available in simulation, which currently has a limited variety of objects and motion. In this work, we demonstrate how large-scale, unlabeled, uncurated real-world data can improve a TAP model with minimal architectural changes, using a selfsupervised student-teacher setup. We demonstrate state-of-the-art performance on the TAP-Vid benchmark surpassing previous results by a wide margin: for example, TAP-Vid-DAVIS performance improves from 61.3% to 67.4%, and TAP-Vid-Kinetics from 57.2% to 62.5%. For visualizations, see our project webpage at https://bootstap.github.io/

Artem Zholus, Carl Doersch, Yi Yang et al.

Tracking Any Point (TAP) in a video is a challenging computer vision problem with many demonstrated applications in robotics, video editing, and 3D reconstruction. Existing methods for TAP rely heavily on complex tracking-specific inductive biases and heuristics, limiting their generality and potential for scaling. To address these challenges, we present TAPNext, a new approach that casts TAP as sequential masked token decoding. Our model is causal, tracks in a purely online fashion, and removes tracking-specific inductive biases. This enables TAPNext to run with minimal latency, and removes the temporal windowing required by many existing state of art trackers. Despite its simplicity, TAPNext achieves a new state-of-the-art tracking performance among both online and offline trackers. Finally, we present evidence that many widely used tracking heuristics emerge naturally in TAPNext through end-to-end training. The TAPNext model and code can be found at https://tap-next.github.io/.

Felipe Codevilla, Jean Gabriel Simard, Ross Goroshin et al.

Recent work has shown that learned image compression strategies can outperform standard hand-crafted compression algorithms that have been developed over decades of intensive research on the rate-distortion trade-off. With growing applications of computer vision, high quality image reconstruction from a compressible representation is often a secondary objective. Compression that ensures high accuracy on computer vision tasks such as image segmentation, classification, and detection therefore has the potential for significant impact across a wide variety of settings. In this work, we develop a framework that produces a compression format suitable for both human perception and machine perception. We show that representations can be learned that simultaneously optimize for compression and performance on core vision tasks. Our approach allows models to be trained directly from compressed representations, and this approach yields increased performance on new tasks and in low-shot learning settings. We present results that improve upon segmentation and detection performance compared to standard high quality JPGs, but with representations that are four to ten times smaller in terms of bits per pixel. Further, unlike naive compression methods, at a level ten times smaller than standard JEPGs, segmentation and detection models trained from our format suffer only minor degradation in performance.

Vincent Dumoulin, Neil Houlsby, Utku Evci et al.

Meta and transfer learning are two successful families of approaches to few-shot learning. Despite highly related goals, state-of-the-art advances in each family are measured largely in isolation of each other. As a result of diverging evaluation norms, a direct or thorough comparison of different approaches is challenging. To bridge this gap, we perform a cross-family study of the best transfer and meta learners on both a large-scale meta-learning benchmark (Meta-Dataset, MD), and a transfer learning benchmark (Visual Task Adaptation Benchmark, VTAB). We find that, on average, large-scale transfer methods (Big Transfer, BiT) outperform competing approaches on MD, even when trained only on ImageNet. In contrast, meta-learning approaches struggle to compete on VTAB when trained and validated on MD. However, BiT is not without limitations, and pushing for scale does not improve performance on highly out-of-distribution MD tasks. In performing this study, we reveal a number of discrepancies in evaluation norms and study some of these in light of the performance gap. We hope that this work facilitates sharing of insights from each community, and accelerates progress on few-shot learning.

Ross Goroshin, Jonathan Tompson, Debidatta Dwibedi

Fully convolutional deep correlation networks are integral components of state-of the-art approaches to single object visual tracking. It is commonly assumed that these networks perform tracking by detection by matching features of the object instance with features of the entire frame. Strong architectural priors and conditioning on the object representation is thought to encourage this tracking strategy. Despite these strong priors, we show that deep trackers often default to tracking by saliency detection - without relying on the object instance representation. Our analysis shows that despite being a useful prior, salience detection can prevent the emergence of more robust tracking strategies in deep networks. This leads us to introduce an auxiliary detection task that encourages more discriminative object representations that improve tracking performance.

Eleni Triantafillou, Tyler Zhu, Vincent Dumoulin et al.

Few-shot classification refers to learning a classifier for new classes given only a few examples. While a plethora of models have emerged to tackle it, we find the procedure and datasets that are used to assess their progress lacking. To address this limitation, we propose Meta-Dataset: a new benchmark for training and evaluating models that is large-scale, consists of diverse datasets, and presents more realistic tasks. We experiment with popular baselines and meta-learners on Meta-Dataset, along with a competitive method that we propose. We analyze performance as a function of various characteristics of test tasks and examine the models' ability to leverage diverse training sources for improving their generalization. We also propose a new set of baselines for quantifying the benefit of meta-learning in Meta-Dataset. Our extensive experimentation has uncovered important research challenges and we hope to inspire work in these directions.

Piotr Mirowski, Razvan Pascanu, Fabio Viola et al.

Learning to navigate in complex environments with dynamic elements is an important milestone in developing AI agents. In this work we formulate the navigation question as a reinforcement learning problem and show that data efficiency and task performance can be dramatically improved by relying on additional auxiliary tasks leveraging multimodal sensory inputs. In particular we consider jointly learning the goal-driven reinforcement learning problem with auxiliary depth prediction and loop closure classification tasks. This approach can learn to navigate from raw sensory input in complicated 3D mazes, approaching human-level performance even under conditions where the goal location changes frequently. We provide detailed analysis of the agent behaviour, its ability to localise, and its network activity dynamics, showing that the agent implicitly learns key navigation abilities.

Ross Goroshin, Michael Mathieu, Yann LeCun

Training deep feature hierarchies to solve supervised learning tasks has achieved state of the art performance on many problems in computer vision. However, a principled way in which to train such hierarchies in the unsupervised setting has remained elusive. In this work we suggest a new architecture and loss for training deep feature hierarchies that linearize the transformations observed in unlabeled natural video sequences. This is done by training a generative model to predict video frames. We also address the problem of inherent uncertainty in prediction by introducing latent variables that are non-deterministic functions of the input into the network architecture.

Junbo Zhao, Michael Mathieu, Ross Goroshin et al.

We present a novel architecture, the "stacked what-where auto-encoders" (SWWAE), which integrates discriminative and generative pathways and provides a unified approach to supervised, semi-supervised and unsupervised learning without relying on sampling during training. An instantiation of SWWAE uses a convolutional net (Convnet) (LeCun et al. (1998)) to encode the input, and employs a deconvolutional net (Deconvnet) (Zeiler et al. (2010)) to produce the reconstruction. The objective function includes reconstruction terms that induce the hidden states in the Deconvnet to be similar to those of the Convnet. Each pooling layer produces two sets of variables: the "what" which are fed to the next layer, and its complementary variable "where" that are fed to the corresponding layer in the generative decoder.

Ross Goroshin, Joan Bruna, Jonathan Tompson et al.

Current state-of-the-art classification and detection algorithms rely on supervised training. In this work we study unsupervised feature learning in the context of temporally coherent video data. We focus on feature learning from unlabeled video data, using the assumption that adjacent video frames contain semantically similar information. This assumption is exploited to train a convolutional pooling auto-encoder regularized by slowness and sparsity. We establish a connection between slow feature learning to metric learning and show that the trained encoder can be used to define a more temporally and semantically coherent metric.

Ross Goroshin, Joan Bruna, Jonathan Tompson et al.

Current state-of-the-art classification and detection algorithms rely on supervised training. In this work we study unsupervised feature learning in the context of temporally coherent video data. We focus on feature learning from unlabeled video data, using the assumption that adjacent video frames contain semantically similar information. This assumption is exploited to train a convolutional pooling auto-encoder regularized by slowness and sparsity. We establish a connection between slow feature learning to metric learning and show that the trained encoder can be used to define a more temporally and semantically coherent metric.

Jonathan Tompson, Ross Goroshin, Arjun Jain et al.

Recent state-of-the-art performance on human-body pose estimation has been achieved with Deep Convolutional Networks (ConvNets). Traditional ConvNet architectures include pooling and sub-sampling layers which reduce computational requirements, introduce invariance and prevent over-training. These benefits of pooling come at the cost of reduced localization accuracy. We introduce a novel architecture which includes an efficient `position refinement' model that is trained to estimate the joint offset location within a small region of the image. This refinement model is jointly trained in cascade with a state-of-the-art ConvNet model to achieve improved accuracy in human joint location estimation. We show that the variance of our detector approaches the variance of human annotations on the FLIC dataset and outperforms all existing approaches on the MPII-human-pose dataset.