While learned video codecs have demonstrated great promise, they have yet to achieve sufficient efficiency for practical deployment. In this work, we propose several novel ideas for learned video compression which allow for improved performance for the low-latency mode (I- and Pframes only) along with a considerable increase in computational efficiency. In this setting, for natural videos our approach compares favorably across the entire R-D curve under metrics PSNR, MS-SSIM and VMAF against all mainstream video standards (H.264, H.265, AV1) and all ML codecs. At the same time, our approach runs at least 5x faster and has fewer parameters than all ML codecs which report these figures. Our contributions include a flexible-rate framework allowing a single model to cover a large and dense range of bitrates, at a negligible increase in computation and parameter count; an efficient backbone optimized for MLbased codecs; and a novel in-loop flow prediction scheme which leverages prior information towards more efficient compression. 140% 120% 100% BD-rate relative to AV1 80% 60% 40% 20% 0%-20% 1.0 Wu et al. Habibian et al. 0.8 Liu et al. 0.6 DVC 0.4 0.2 Ours 0.0 100 200 400 Encode time (ms) 800 0.0 Wu et al. Liu et al. Habibian et al. DVC Ours 50 0.2 200 0.4 600 105 0.6 106 0.8 Decode time (ms) 107 1.0 Figure 1: BD-Rate for ML-based codecs relative to AV1 as a function of encode/decode time on HD 1080 videos [42, 14, 29, 27] (UVGdataset, PSNR metric). Our approach reduces the BD-rate by 54% relative to the current fastest MLcodecwhichreportsspeed[27], whilerunning5xfaster. We benchmark our method, which we call ELF-VC (Eff icient, Learned and Flexible Video Coding) on popular video test sets UVG and MCL-JCV under metrics PSNR, MS-SSIM and VMAF. For example, on UVG under PSNR, it reduces the BD-rate by 44% against H.264, 26% against H.265, 15% against AV1, and 35% against the current best ML codec. At the same time, on an NVIDIA Titan V GPU our approach encodes/decodes VGA at 49/91 FPS, HD 720 at 19/35 FPS, and HD 1080 at 10/18 FPS.

We present a new algorithm for video coding, learned end-to-end for the low-latency mode. In this setting, our approach outperforms all existing video codecs across nearly the entire bitrate range. To our knowledge, this is the first ML-based method to do so. We evaluate our approach on standard video compression test sets of varying resolutions, and benchmark against all mainstream commercial codecs, in the low-latency mode. On standard-definition videos, relative to our algorithm, HEVC/H.265, AVC/H.264 and VP9 typically produce codes up to 60% larger. On high-definition 1080p videos, H.265 and VP9 typically produce codes up to 20% larger, and H.264 up to 35% larger. Furthermore, our approach does not suffer from blocking artifacts and pixelation, and thus produces videos that are more visually pleasing. We propose two main contributions. The first is a novel architecture for video compression, which (1) generalizes motion estimation to perform any learned compensation beyond simple translations, (2) rather than strictly relying on previously transmitted reference frames, maintains a state of arbitrary information learned by the model, and (3) enables jointly compressing all transmitted signals (such as optical flow and residual). Secondly, we present a framework for ML-based spatial rate control — a mechanism for assigning variable bitrates across space for each frame. This is a critical component for video coding, which to our knowledge had not been developed within a machine learning setting.

We present a machine learning-based approach to lossy image compression which outperforms all existing codecs, while running in real-time. Our algorithm typically produces files 2.5 times smaller than JPEG and JPEG 2000, 2 times smaller than WebP, and 1.7 times smaller than BPG on datasets of generic images across all quality levels. At the same time, our codec is designed to be lightweight and deployable: for example, it can encode or decode the Kodak dataset in around 10ms per image on GPU. Our architecture is an autoencoder featuring pyramidal analysis, an adaptive coding module, and regularization of the expected codelength. We also supplement our approach with adversarial training specialized towards use in a compression setting: this enables us to produce visually pleasing reconstructions for very low bitrates.

This paper aims to classify and locate objects accurately and efficiently, without using bounding box annotations. It is challenging as objects in the wild could appear at arbitrary locations and in different scales. We propose a novel classification architecture ProNet based on convolutional neural networks. It uses computationally efficient neural networks to propose image regions that are likely to contain objects, and applies more powerful but slower networks on the proposed regions. The basic building block is a multi-scale fully-convolutional network which assigns object confidence scores to boxes at different locations and scales. We show that such networks can be trained effectively using image-level annotations, and can be connected into cascades or trees for efficient object classification. ProNet outperforms previous state-of-the-art on PASCAL VOC 2012 and MS COCO datasets for object classification and point-based localization.

Over the last few years deep learning methods have emerged as one of the most prominent approaches for video analysis. However, so far their most successful applications have been in the area of video classification and detection, i.e., problems involving the prediction of a single class label or a handful of output variables per video. Furthermore, while deep networks are commonly recognized as the best models to use in these domains, there is a widespread perception that in order to yield successful results they often require time-consuming architecture search, manual tweaking of parameters and computationally intensive preprocessing or post-processing methods. In this paper we challenge these views by presenting a deep 3D convolutional architecture trained end to end to perform voxel-level prediction, i.e., to output a variable at every voxel of the video. Most importantly, we show that the same exact architecture can be used to achieve competitive results on three widely different voxel-prediction tasks: video semantic segmentation, optical flow estimation, and video coloring. The three networks learned on these problems are trained from raw video without any form of preprocessing and their outputs do not require post-processing to achieve outstanding performance. Thus, they offer an efficient alternative to traditional and much more computationally expensive methods in these video domains.

Distance metric learning (DML) approaches learn a transformation to a representation space where distance is in correspondence with a predefined notion of similarity. While such models offer a number of compelling benefits, it has been difficult for these to compete with modern classification algorithms in performance and even in feature extraction. In this work, we propose a novel approach explicitly designed to address a number of subtle yet important issues which have stymied earlier DML algorithms. It maintains an explicit model of the distributions of the different classes in representation space. It then employs this knowledge to adaptively assess similarity, and achieve local discrimination by penalizing class distribution overlap. We demonstrate the effectiveness of this idea on several tasks. Our approach achieves state-of-the-art classification results on a number of fine-grained visual recognition datasets, surpassing the standard softmax classifier and outperforming triplet loss by a relative margin of 30-40%. In terms of computational performance, it alleviates training inefficiencies in the traditional triplet loss, reaching the same error in 5-30 times fewer iterations. Beyond classification, we further validate the saliency of the learnt representations via their attribute concentration and hierarchy recovery properties, achieving 10-25% relative gains on the softmax classifier and 25-50% on triplet loss in these tasks.

With the widespread availability of cellphones and cameras that have GPS capabilities, it is common for images being uploaded to the Internet today to have GPS coordinates associated with them. In addition to research that tries to predict GPS coordinates from visual features, this also opens up the door to problems that are conditioned on the availability of GPS coordinates. In this work, we tackle the problem of performing image classification with location context, in which we are given the GPS coordinates for images in both the train and test phases. We explore different ways of encoding and extracting features from the GPS coordinates, and show how to naturally incorporate these features into a Convolutional Neural Network (CNN), the current state-of-the-art for most image classification and recognition problems. We also show how it is possible to simultaneously learn the optimal pooling radii for a subset of our features within the CNN framework. To evaluate our model and to help promote research in this area, we identify a set of location-sensitive concepts and annotate a subset of the Yahoo Flickr Creative Commons 100M dataset that has GPS coordinates with these concepts, which we make publicly available. By leveraging location context, we are able to achieve almost a 7% gain in mean average precision.

We propose a simple, yet effective approach for spatiotemporal feature learning using deep 3-dimensional convolutional networks (3D ConvNets) trained on a large scale supervised video dataset. Our findings are three-fold: 1) 3D ConvNets are more suitable for spatiotemporal feature learning compared to 2D ConvNets; 2) A homogeneous architecture with small 3x3x3 convolution kernels in all layers is among the best performing architectures for 3D ConvNets; and 3) Our learned features, namely C3D (Convolutional 3D), with a simple linear classifier outperform state-of-the-art methods on 4 different benchmarks and are comparable with current best methods on the other 2 benchmarks. In addition, the features are compact: achieving 52.8% accuracy on UCF101 dataset with only 10 dimensions and also very efficient to compute due to the fast inference of ConvNets. Finally, they are conceptually very simple and easy to train and use.

We explore the task of recognizing peoples' identities in photo albums in an unconstrained setting. To facilitate this, we introduce the new People In Photo Albums (PIPA) dataset, consisting of over 60000 instances of over 2000 individuals collected from public Flickr photo albums. With only about half of the person images containing a frontal face, the recognition task is very challenging due to the large variations in pose, clothing, camera viewpoint, image resolution and illumination. We propose the Pose Invariant PErson Recognition (PIPER) method, which accumulates the cues of poselet-level person recognizers trained by deep convolutional networks to discount for the pose variations, combined with a face recognizer and a global recognizer. Experiments on three different settings confirm that in our unconstrained setup PIPER significantly improves on the performance of DeepFace, which is one of the best face recognizers as measured on the LFW dataset.

This paper addresses the problem of clustering a very large number of photos (i.e. hundreds of millions a day) in a stream into millions of clusters. This is particularly important as the popularity of photo sharing websites, such as Facebook, Google, and Instagram. Given large number of photos available online, how to efficiently organize them is an open problem. To address this problem, we propose to cluster the binary hash codes of a large number of photos into binary cluster centers. We present a fast binary k-means algorithm that works directly on the similarity-preserving hashes of images and clusters them into binary centers on which we can build hash indexes to speedup computation. The proposed method is capable of clustering millions of photos on a single machine in a few minutes. We show that this approach is usually several magnitude faster than standard k-means and produces comparable clustering accuracy. In addition, we propose an online clustering method based on binary k-means that is capable of clustering large photo stream on a single machine, and show applications to spam detection and trending photo discovery.

The availability of large labeled datasets has allowed Convolutional Network models to achieve impressive recognition results. However, in many settings manual annotation of the data is impractical; instead our data has noisy labels, i.e. there is some freely available label for each image which may or may not be accurate. In this paper, we explore the performance of discriminatively-trained Convnets when trained on such noisy data. We introduce an extra noise layer into the network which adapts the network outputs to match the noisy label distribution. The parameters of this noise layer can be estimated as part of the training process and involve simple modifications to current training infrastructures for deep networks. We demonstrate the approaches on several datasets, including large scale experiments on the ImageNet classification benchmark.

We present a new dataset with the goal of advancing the state-of-the-art in object recognition by placing the question of object recognition in the context of the broader question of scene understanding. This is achieved by gathering images of complex everyday scenes containing common objects in their natural context. Objects are labeled using per-instance segmentations to aid in precise object localization. Our dataset contains photos of 91 objects types that would be easily recognizable by a 4 year old. With a total of 2.5 million labeled instances in 328k images, the creation of our dataset drew upon extensive crowd worker involvement via novel user interfaces for category detection, instance spotting and instance segmentation. We present a detailed statistical analysis of the dataset in comparison to PASCAL, ImageNet, and SUN. Finally, we provide baseline performance analysis for bounding box and segmentation detection results using a Deformable Parts Model.

We propose a method for inferring human attributes (such as gender, hair style, clothes style, expression, action) from images of people under large variation of viewpoint, pose, appearance, articulation and occlusion. Convolutional Neural Nets (CNN) have been shown to perform very well on large scale object recognition problems. In the context of attribute classification, however, the signal is often subtle and it may cover only a small part of the image, while the image is dominated by the effects of pose and viewpoint. Discounting for pose variation would require training on very large labeled datasets which are not presently available. Part-based models, such as poselets and DPM have been shown to perform well for this problem but they are limited by flat low-level features. We propose a new method which combines part-based models and deep learning by training pose-normalized CNNs. We show substantial improvement vs. state-of-the-art methods on challenging attribute classification tasks in unconstrained settings. Experiments confirm that our method outperforms both the best part-based methods on this problem and conventional CNNs trained on the full bounding box of the person.

Poselets have been used in a variety of computer vision tasks, such as detection, segmentation, action classification, pose estimation and action recognition, often achieving state-of-the-art performance. Poselet evaluation, however, is computationally intensive as it involves running thousands of scanning window classifiers. We present an algorithm for training a hierarchical cascade of part-based detectors and apply it to speed up poselet evaluation. Our cascade hierarchy leverages common components shared across poselets. We generate a family of cascade hierarchies, including trees that grow logarithmically on the number of poselet classifiers. Our algorithm, under some reasonable assumptions, finds the optimal tree structure that maximizes speed for a given target detection rate. We test our system on the PASCAL dataset and show an order of magnitude speedup at less than 1% loss in AP.

We address the problem of detecting people in natural scenes using a part approach based on poselets. We propose a bootstrapping method that allows us to collect millions of weakly labeled examples for each poselet type. We use these examples to train a Convolutional Neural Net to discriminate different poselet types and separate them from the background class. We then use the trained CNN as a way to represent poselet patches with a Pose Discriminative Feature (PDF) vector -- a compact 256-dimensional feature vector that is effective at discriminating pose from appearance. We train the poselet model on top of PDF features and combine them with object-level CNNs for detection and bounding box prediction. The resulting model leads to state-of-the-art performance for human detection on the PASCAL datasets.

We propose a novel approach for human pose estimation in real-world cluttered scenes, and focus on the challenging problem of predicting the pose of both arms for each person in the image. For this purpose, we build on the notion of poselets and train highly discriminative classifiers to differentiate among arm configurations, which we call armlets. We propose a rich representation which, in addition to standard HOG features, integrates the information of strong contours, skin color and contextual cues in a principled manner. Unlike existing methods, we evaluate our approach on a large subset of images from the PASCAL VOC detection dataset, where critical visual phenomena, such as occlusion, truncation, multiple instances and clutter are the norm. Our approach outperforms Yang and Ramanan, the state-of-the-art technique, with an improvement from 29.0% to 37.5% PCP accuracy on the arm keypoint prediction task, on this new pose estimation dataset.

We address the problem of interactive facial feature localization from a single image. Our goal is to obtain an accurate segmentation of facial features on high-resolution images under a variety of pose, expression, and lighting conditions. Although there has been significant work in facial feature localization, we are addressing a new application area, namely to facilitate intelligent high-quality editing of portraits, that brings requirements not met by existing methods. We propose an improvement to the Active Shape Model that allows for greater independence among the facial components and improves on the appearance fitting step by introducing a Viterbi optimization process that operates along the facial contours. Despite the improvements, we do not expect perfect results in all cases. We therefore introduce an interaction model whereby a user can efficiently guide the algorithm towards a precise solution. We introduce the Helen Facial Feature Dataset consisting of annotated portrait images gathered from Flickr that are more diverse and challenging than currently existing datasets. We present experiments that compare our automatic method to published results, and also a quantitative evaluation of the effectiveness of our interactive method.

We address the problem of segmenting and recognizing objects in real world images, focusing on challenging articulated categories such as humans and other animals. For this purpose, we propose a novel design for region-based object detectors that integrates efficiently top-down information from scanning-windows part models and global appearance cues. Our detectors produce class-specific scores for bottom-up regions, and then aggregate the votes of multiple overlapping candidates through pixel classification. We evaluate our approach on the PASCAL segmentation challenge, and report competitive performance with respect to current leading techniques. On VOC2010, our method obtains the best results in 6/20 categories and the highest performance on articulated objects.

We address the problem of editing facial expression in video, such as exaggerating, attenuating or replacing the expression with a different one in some parts of the video. To achieve this we develop a tensor-based 3D face geometry reconstruction method, which fits a 3D model for each video frame, with the constraint that all models have the same identity and requiring temporal continuity of pose and expression. With the identity constraint, the differences between the underlying 3D shapes capture only changes in expression and pose. We show that various expression editing tasks in video can be achieved by combining face reordering with face warping, where the warp is induced by projecting differences in 3D face shapes into the image plane. Analogously, we show how the identity can be manipulated while fixing expression and pose. Experimental results show that our method can effectively edit expressions and identity in video in a temporally-coherent way with high fidelity.

The explosive growth in image sharing via social networks has produced exciting opportunities for the computer vision community in areas including face, text, product and scene recognition. In this work we turn our attention to group photos of people and ask the question: what can we determine about the social subculture or urban tribe to which these people belong? To this end, we propose a framework employing low- and mid-level features to capture the visual attributes distinctive to a variety of urban tribes. We proceed in a semi-supervised manner, employing a metric that allows us to extrapolate from a small number of pairwise image similarities to induce a set of groups that visually correspond to familiar urban tribes such as biker, hipster or goth. Automatic recognition of such information in group photos offers the potential to improve recommendation services, context sensitive advertising and other social analysis applications. We present promising preliminary experimental results that demonstrate our ability to categorize group photos in a socially meaningful manner

We propose a method for recognizing attributes, such as the gender, hair style and types of clothes of people under large variation in viewpoint, pose, articulation and occlusion typical of personal photo album images. Robust attribute classifiers under such conditions must be invariant to pose, but inferring the pose in itself is a challenging problem. We use a part-based approach based on poselets. Our parts implicitly decompose the aspect (the pose and viewpoint). We train attribute classifiers for each such aspect and we combine them together in a discriminative model. We propose a new dataset of 8000 people with annotated attributes. Our method performs very well on this dataset, significantly outperforming a baseline built on the spatial pyramid match kernel method. On gender recognition we outperform a commercial face recognition system.

We study the challenging problem of localizing and classifying category-specific object contours in real world images. For this purpose, we present a simple yet effective method for combining generic object detectors with bottomup contours to identify object contours. We also provide a principled way of combining information from different part detectors and across categories. In order to study the problem and evaluate quantitatively our approach, we present a dataset of semantic exterior boundaries on more than 20, 000 object instances belonging to 20 categories, using the images from the VOC2011 PASCAL challenge

We present a distributed representation of pose and appearance of people called the “poselet activation vector”. First we show that this representation can be used to estimate the pose of people defined by the 3D orientations of the head and torso in the challenging PASCAL VOC 2010 person detection dataset. Our method is robust to clutter, aspect and viewpoint variation and works even when body parts like faces and limbs are occluded or hard to localize. We combine this representation with other sources of information like interaction with objects and other people in the image and use it for action recognition. We report competitive results on the PASCAL VOC 2010 static image action classification challenge

In this paper, we propose techniques to make use of two complementary bottom-up features, image edges and texture patches, to guide top-down object segmentation towards higher precision. We build upon the part-based poselet detector, which can predict masks for numerous parts of an object. For this purpose we extend poselets to 19 other categories apart from person. We non-rigidly align these part detections to potential object contours in the image, both to increase the precision of the predicted object mask and to sort out false positives. We spatially aggregate object information via a variational smoothing technique while ensuring that object regions do not overlap. Finally, we propose to refine the segmentation based on self-similarity de- fined on small image patches. We obtain competitive results on the challenging Pascal VOC benchmark. On four classes we achieve the best numbers to-date.

We address the classic problems of detection and segmentation using a part based detector that operates on a novel part, which we refer to as a poselet. Poselets are tightly clustered in both appearance space (and thus are easy to detect) as well as in configuration space (and thus are helpful for localization and segmentation). We demonstrate poselets are effective for detection, pose extraction, segmentation, action/pose estimation and attribute classification. Poselet construction requires extra annotations beyond the object bounds. To train poselets we have created H3D (Humans in 3D) - a dataset of 1200+ person annotations. The annotations include the joints, the extracted 3D pose, keypoint visibility and region labels. We have also annotated the people in the training and validation sets of PASCAL VOC 2009. Our poselet classifier achieves state-of-the-art results for the person category on PASCAL VOC 2007, 2008, 2009 and 2010 as well as on our dataset, H3D.

Bourdev and Malik (ICCV 09) introduced a new notion of parts, poselets, constructed to be tightly clustered both in the configuration space of keypoints, as well as in the appearance space of image patches. In this paper we develop a new algorithm for detecting people using poselets. Unlike that work which used 3D annotations of keypoints, we use only 2D annotations which are much easier for naive human annotators. The main algorithmic contribution is in how we use the pattern of poselet activations. Individual poselet activations are noisy, but considering the spatial context of each can provide vital disambiguating information, just as object detection can be improved by considering the detection scores of nearby objects in the scene. This can be done by training a two-layer feed-forward network with weights set using a max margin technique. The refined poselet activations are then clustered into mutually consistent hypotheses where consistency is based on empirically determined spatial keypoint distributions. Finally, bounding boxes are predicted for each person hypothesis and shape masks are aligned to edges in the image to provide a segmentation. To the best of our knowledge, the resulting system is the current best performer on the task of people detection and segmentation with an average precision of 47.8% and 40.5% respectively on PASCAL VOC 2009.

We address the classic problems of detection, segmentation and pose estimation of people in images with a novel definition of a part, a poselet. We postulate two criteria (1) It should be easy to find a poselet given an input image (2) it should be easy to localize the 3D configuration of the person conditioned on the detection of a poselet. To permit this we have built a new dataset, H3D, of annotations of humans in 2D photographs with 3D joint information, inferred using anthropometric constraints. This enables us to implement a data-driven search procedure for finding poselets that are tightly clustered in both 3D joint configuration space as well as 2D image appearance. The algorithm discovers poselets that correspond to frontal and profile faces, pedestrians, head and shoulder views, among others. Each poselet provides examples for training a linear SVM classifier which can then be run over the image in a multiscale scanning mode. The outputs of these poselet detectors can be thought of as an intermediate layer of nodes, on top of which one can run a second layer of classification or regression. We show how this permits detection and localization of torsos or keypoints such as left shoulder, nose, etc. Experimental results show that we obtain state of the art performance on people detection in the PASCAL VOC 2007 challenge, among other datasets. We are making publicly available both the H3D dataset as well as the poselet parameters for use by other researchers.

We describe a method for training object detectors using a generalization of the cascade architecture, which results in a detection rate and speed comparable to that of the best published detectors while allowing for easier training and a detector with fewer features. In addition, the method allows for quickly calibrating the detector for a target detection rate, false positive rate or speed. One important advantage of our method is that it enables systematic exploration of the ROC Surface, which characterizes the trade-off between accuracy and speed for a given classifier.