Lossy Image Compression with Conditional Diffusion Models

Diffusion models are a new class of generative models that mark a milestone in high-quality image generation while relying on solid probabilistic principles. This makes them promising candidate models for neural image compression. This paper outlines an end-to-end optimized framework based on a conditional diffusion model for image compression. Besides latent variables inherent to the diffusion process, the model introduces an additional per-instance "content" latent variable to condition the denoising process. Upon decoding, the diffusion process conditionally generates/reconstructs an image using ancestral sampling. Our experiments show that this approach outperforms one of the best-performing conventional image codecs (BPG) and one neural codec on two compression benchmarks, where we focus on rate-perception tradeoffs. Qualitatively, our approach shows fewer decompression artifacts than the classical approach.

Diffusion Probabilistic Modeling for Video Generation

Denoising diffusion probabilistic models are a promising new class of generative models that are competitive with GANs on perceptual metrics. In this paper, we explore their potential for sequentially generating video. Inspired by recent advances in neural video compression, we use denoising diffusion models to stochastically generate a residual to a deterministic next-frame prediction. We compare this approach to two sequential VAE and two GAN baselines on four datasets, where we test the generated frames for perceptual quality and forecasting accuracy against ground truth frames. We find significant improvements in terms of perceptual quality on all data and improvements in terms of frame forecasting for complex high-resolution videos.

SC2: Supervised Compression for Split Computing

Split computing distributes the execution of a neural network (e.g., for a classification task) between a mobile device and a more powerful edge server. A simple alternative to splitting the network is to carry out the supervised task purely on the edge server while compressing and transmitting the full data, and most approaches have barely outperformed this baseline. This paper proposes a new approach for discretizing and entropy-coding intermediate feature activations to efficiently transmit them from the mobile device to the edge server. We show that a efficient splittable network architecture results from a three-way tradeoff between (a) minimizing the computation on the mobile device, (b) minimizing the size of the data to be transmitted, and (c) maximizing the model's prediction performance. We propose an architecture based on this tradeoff and train the splittable network and entropy model in a knowledge distillation framework. In an extensive set of experiments involving three vision tasks, three datasets, nine baselines, and more than 180 trained models, we show that our approach improves supervised rate-distortion tradeoffs while maintaining a considerably smaller encoder size. We also release sc2bench, an installable Python package, to encourage and facilitate future studies on supervised compression for split computing (SC2).

Vision-Guided Quadrupedal Locomotion in the Wild with Multi-Modal Delay Randomization

Developing robust vision-guided controllers for quadrupedal robots in complex environments, with various obstacles, dynamical surroundings and uneven terrains, is very challenging. While Reinforcement Learning (RL) provides a promising paradigm for agile locomotion skills with vision inputs in simulation, it is still very challenging to deploy the RL policy in the real world. Our key insight is that aside from the discrepancy in the domain gap, in visual appearance between the simulation and the real world, the latency from the control pipeline is also a major cause of difficulty. In this paper, we propose Multi-Modal Delay Randomization (MMDR) to address this issue when training RL agents. Specifically, we simulate the latency of real hardware by using past observations, sampled with randomized periods, for both proprioception and vision. We train the RL policy for end-to-end control in a physical simulator without any predefined controller or reference motion, and directly deploy it on the real A1 quadruped robot running in the wild. We evaluate our method in different outdoor environments with complex terrains and obstacles. We demonstrate the robot can smoothly maneuver at a high speed, avoid the obstacles, and show significant improvement over the baselines. Our project page with videos is at https://mehooz.github.io/mmdr-wild/.

DexMV: Imitation Learning for Dexterous Manipulation from Human Videos

While we have made significant progress on understanding hand-object interactions in computer vision, it is still very challenging for robots to perform complex dexterous manipulation. In this paper, we propose a new platform and pipeline, DexMV (Dexterous Manipulation from Videos), for imitation learning to bridge the gap between computer vision and robot learning. We design a platform with: (i) a simulation system for complex dexterous manipulation tasks with a multi-finger robot hand and (ii) a computer vision system to record large-scale demonstrations of a human hand conducting the same tasks. In our new pipeline, we extract 3D hand and object poses from the videos, and convert them to robot demonstrations via motion retargeting. We then apply and compare multiple imitation learning algorithms with the demonstrations. We show that the demonstrations can indeed improve robot learning by a large margin and solve the complex tasks which reinforcement learning alone cannot solve. Project page with video: https://yzqin.github.io/dexmv

Supervised Compression for Resource-constrained Edge Computing Systems

There has been much interest in deploying deep learning algorithms on low-powered devices, including smartphones, drones, and medical sensors. However, full-scale deep neural networks are often too resource-intensive in terms of energy and storage. As a result, the bulk part of the machine learning operation is therefore often carried out on an edge server, where the data is compressed and transmitted. However, compressing data (such as images) leads to transmitting information irrelevant to the supervised task. Another popular approach is to split the deep network between the device and the server while compressing intermediate features. To date, however, such split computing strategies have barely outperformed the aforementioned naive data compression baselines due to their inefficient approaches to feature compression. This paper adopts ideas from knowledge distillation and neural image compression to compress intermediate feature representations more efficiently. Our supervised compression approach uses a teacher model and a student model with a stochastic bottleneck and learnable prior for entropy coding. We compare our approach to various neural image and feature compression baselines in three vision tasks and found that it achieves better supervised rate-distortion performance while also maintaining smaller end-to-end latency. We furthermore show that the learned feature representations can be tuned to serve multiple downstream tasks.

Insights from Generative Modeling for Neural Video Compression

While recent machine learning research has revealed connections between deep generative models such as VAEs and rate-distortion losses used in learned compression, most of this work has focused on images. In a similar spirit, we view recently proposed neural video coding algorithms through the lens of deep autoregressive and latent variable modeling. We present recent neural video codecs as instances of a generalized stochastic temporal autoregressive transform, and propose new avenues for further improvements inspired by normalizing flows and structured priors. We propose several architectures that yield state-of-the-art video compression performance on full-resolution video and discuss their tradeoffs and ablations. In particular, we propose (i) improved temporal autoregressive transforms, (ii) improved entropy models with structured and temporal dependencies, and (iii) variable bitrate versions of our algorithms. Since our improvements are compatible with a large class of existing models, we provide further evidence that the generative modeling viewpoint can advance the neural video coding field.

Learning Vision-Guided Quadrupedal Locomotion End-to-End with Cross-Modal Transformers

We propose to address quadrupedal locomotion tasks using Reinforcement Learning (RL) with a Transformer-based model that learns to combine proprioceptive information and high-dimensional depth sensor inputs. While learning-based locomotion has made great advances using RL, most methods still rely on domain randomization for training blind agents that generalize to challenging terrains. Our key insight is that proprioceptive states only offer contact measurements for immediate reaction, whereas an agent equipped with visual sensory observations can learn to proactively maneuver environments with obstacles and uneven terrain by anticipating changes in the environment many steps ahead. In this paper, we introduce LocoTransformer, an end-to-end RL method for quadrupedal locomotion that leverages a Transformer-based model for fusing proprioceptive states and visual observations. We evaluate our method in challenging simulated environments with different obstacles and uneven terrain. We show that our method obtains significant improvements over policies with only proprioceptive state inputs, and that Transformer-based models further improve generalization across environments. Our project page with videos is at https://RchalYang.github.io/LocoTransformer .