We present Pippo, a generative model capable of producing 1K resolution dense turnaround videos of a person from a single casually clicked photo. Pippo is a multi-view diffusion transformer and does not require any additional inputs — e.g., a fitted parametric model or camera parameters of the input image. We pre-train Pippo on 3B human images without captions, and conduct multi-view mid-training and post-training on studio captured humans. During mid-training, to quickly absorb the studio dataset, we denoise several (upto 48) views at low-resolution, and encode target cameras coarsely using a shallow MLP. During post-training, we denoise fewer views at high-resolution and use pixel-aligned controls (e.g., Spatial anchor and Plucker rays) to enable 3D consistent generations. At inference, we propose an attention biasing technique that allows Pippo to simultaneously generate greater than 5× as many views as seen during training. Finally, we also introduce an improved metric to evaluate 3D consistency of multi-view generations, and show that Pippo outperforms existing works on multi-view human generation from a single image.

To build photorealistic avatars that users can embody, human modelling must be complete (cover the full body), driveable (able to reproduce the current motion and appearance from the user), and generalizable (i.e., easily adaptable to novel identities).Towards these goals, paired captures, that is, captures of the same subject obtained from systems of diverse quality and availability, are crucial.However, paired captures are rarely available to researchers outside of dedicated industrial labs: Codec Avatar Studio is our proposal to close this gap.Towards generalization and driveability, we introduce a dataset of 256 subjects captured in two modalities: high resolution multi-view scans of their heads, and video from the internal cameras of a headset.Towards completeness, we introduce a dataset of 4 subjects captured in eight modalities: high quality relightable multi-view captures of heads and hands, full body multi-view captures with minimal and regular clothes, and corresponding head, hands and body phone captures.Together with our data, we also provide code and pre-trained models for different state-of-the-art human generation models.Our datasets and code are available at https://github.com/facebookresearch/ava-256 and https://github.com/facebookresearch/goliath.

Compressing large neural networks is an important step for their deployment in resource-constrained computational platforms. In this context, vector quantization is an appealing framework that expresses multiple parameters using a single code, and has recently achieved state-of-the-art network compression on a range of core vision and natural language processing tasks. Key to the success of vector quantization is deciding which parameter groups should be compressed together. Previous work has relied on heuristics that group the spatial dimension of individual convolutional filters, but a general solution remains unaddressed. This is desirable for pointwise convolutions (which dominate modern architectures), linear layers (which have no notion of spatial dimension), and convolutions (when more than one filter is compressed to the same codeword). In this paper we make the observation that the weights of two adjacent layers can be permuted while expressing the same function. We then establish a connection to rate-distortion theory and search for permutations that result in networks that are easier to compress. Finally, we rely on an annealed quantization algorithm to better compress the network and achieve higher final accuracy. We show results on image classification, object detection, and segmentation, reducing the gap with the uncompressed model by 40 to 70% w.r.t. the current state of the art. All our experiments can be reproduced using the code at https://github.com/uber-research/permute-quantize-finetune.

Over the last few years, we have witnessed tremendous progress on many subtasks of autonomous driving including perception, motion forecasting, and motion planning. However, these systems often assume that the car is accurately localized against a high-definition map. In this paper we question this assumption, and investigate the issues that arise in state-of-the-art autonomy stacks under localization error. Based on our observations, we design a system that jointly performs perception, prediction, and localization. Our architecture is able to reuse computation between the three tasks, and is thus able to correct localization errors efficiently. We show experiments on a large-scale autonomy dataset, demonstrating the efficiency and accuracy of our proposed approach.

We are interested in understanding whether retrieval-based localization approaches are good enough in the context of self-driving vehicles. Towards this goal, we introduce Pit30M, a new image and LiDAR dataset with over 30 million frames, which is 10 to 100 times larger than those used in previous work. Pit30M is captured under diverse conditions (i.e., season, weather, time of the day, traffic), and provides accurate localization ground truth. We also automatically annotate our dataset with historical weather and astronomical data, as well as with image and LiDAR semantic segmentation as a proxy measure for occlusion. We benchmark multiple existing methods for image and LiDAR retrieval and, in the process, introduce a simple, yet effective convolutional network-based LiDAR retrieval method that is competitive with the state of the art. Our work provides, for the first time, a benchmark for sub-metre retrieval-based localization at city scale.The dataset, additional experimental results, as well as more information about the sensors, calibration, and metadata, are available on the project website: https://uber.com/atg/datasets/pit30m.

One of the main difficulties of scaling current localization systems to large environments is the on-board storage required for the maps. In this paper we propose to learn to compress the map representation such that it is optimal for the localization task. As a consequence, higher compression rates can be achieved without loss of localization accuracy when compared to standard coding schemes that optimize for reconstruction, thus ignoring the end task. Our experiments show that it is possible to learn a task-specific compression which reduces storage requirements by two orders of magnitude over general-purpose codecs such as WebP without sacrificing performance.

Human motion modelling is a classical problem at the intersection of graphics and computer vision, with applications spanning human-computer interaction, motion synthesis, and motion prediction for virtual and augmented reality. Following the success of deep learning methods in several computer vision tasks, recent work has focused on using deep recurrent neural networks (RNNs) to model human motion, with the goal of learning time-dependent representations that perform tasks such as short-term motion prediction and long-term human motion synthesis. We examine recent work, with a focus on the evaluation methodologies commonly used in the literature, and show that, surprisingly, state of the art performance can be achieved by a simple baseline that does not attempt to model motion at all. We investigate this result, and analyze recent RNN methods by looking at the architectures, loss functions, and training procedures used in state-of-the-art approaches. We propose three changes to the standard RNN models typically used for human motion, which results in a simple and scalable RNN architecture that obtains state-of-the-art performance on human motion prediction.

Following the success of deep convolutional networks, state-of-the-art methods for 3d human pose estimation have focused on deep end-to-end systems that predict 3d joint locations given raw image pixels. Despite their excellent performance, it is often not easy to understand whether their remaining error stems from a limited 2dpose (visual) understanding, or from a failure to map 2d poses into 3dimensional positions. With the goal of understanding these sources of error, we set out to build a system that given 2d joint locations predicts 3d positions. Much to our surprise, we have found that, with current technology, “lifting” ground truth 2djoint locations to 3d space is a task that can be solved with a remarkably low error rate: a relatively simple deep feedforward network outperforms the best reported result by about 30% on Human3.6M, the largest publicly available 3d pose estimation benchmark. Furthermore, training our system on the output of an off-the-shelf state-of-the-art 2d detector (i.e., using images as input) yields state of the art results - this includes an array of systems that have been trained end-to-end specifically for this task. Our results indicate that a large portion of the error of modern deep 3d pose estimation systems stems from their visual analysis, and suggests directions to further advance the state of the art in 3d human pose estimation.

We present a novel and scalable approach for retrieval and flexible alignment of 3d human motion examples given a video query. Our method efficiently searches a large set of motion capture (mocap) files accounting for speed variations in motion. To align a short video clip with a part of a longer mocap sequence, we experiment with different feature representations comparable across the two modalities. We also evaluate two different Dynamic Time Warping (DTW) approaches that allow sub-sequence matching and suggest additional local constraints for a smooth alignment. Finally, to quantify video-based mocap retrieval, we introduce a benchmark providing a novel set of per-frame action labels for 2 000 files of the CMU-mocap dataset, as well as a collection of realistic video queries taken from YouTube. Our experiments show that temporal flexibility is not only required for the correct alignment of pose and motion, but it also improves the retrieval accuracy.

We explore a novel paradigm in learning binary codes for large-scale image retrieval applications. Instead of learning a single globally optimal quantization model as in previous approaches, we encode the database points in a data-specific manner using a bank of quantization models. Each individual database point selects the quantization model that minimizes its individual quantization error. We apply the idea of a bank of quantization models to data independent and data-driven hashing methods for learning binary codes, obtaining state-of-the-art performance on three benchmark datasets.

We describe a new approach to transfer knowledge across views for action recognition by using examples from a large collection of unlabelled mocap data. We achieve this by directly matching purely motion based features from videos to mocap. Our approach recovers 3D pose sequences without performing any body part tracking. We use these matches to generate multiple motion projections and thus add view invariance to our action recognition model. We also introduce a closed form solution for approximate non-linear Circulant Temporal Encoding (nCTE), which allows us to efficiently perform the matches in the frequency domain. We test our approach on the challenging unsupervised modality of the IXMAS dataset, and use publicly available motion capture data for matching. Without any additional annotation effort, we are able to significantly outperform the current state of the art.

Nearest Neighbour Search in high-dimensional spaces is a common problem in Computer Vision. Although no algorithm better than linear search is known, approximate algorithms are commonly used to tackle this problem. The drawback of using such algorithms is that their performance depends highly on parameter tuning. While this process can be automated using standard empirical optimization techniques, tuning is still time-consuming. In this paper, we propose to use Empirical Hardness Models to reduce the number of parameter configurations that Bayesian Optimization has to try, speeding up the optimization process. Evaluation on standard benchmarks of SIFT and GIST descriptors shows the viability of our approach.