WACV 2021 Papers

Wearable cameras allow to collect images and videos of humans interacting with the world. While human-object interactions have been thoroughly investigated in third person vision, the problem has been understudied in egocentric settings and in industrial scenarios. To fill this gap, we introduce MECCANO, the first dataset of egocentric videos to study human-object interactions in industrial-like settings. MECCANO has been acquired by 20 participants who were asked to build a motorbike model, for which they had to interact with tiny objects and tools. The dataset has been explicitly labeled for the task of recognizing human-object interactions from an egocentric perspective. Specifically, each interaction has been labeled both temporally (with action segments) and spatially (with active object bounding boxes). With the proposed dataset, we investigate four different tasks including 1) action recognition, 2) active object detection, 3) active object recognition and 4) egocentric human-object interaction detection, which is a revisited version of the standard human-object interaction detection task. Baseline results show that the MECCANO dataset is a challenging benchmark to study egocentric human-object interactions in industrial-like scenarios. We publicy release the dataset at https://iplab.dmi.unict.it/MECCANO/.

Humor is a very important communication tool; yet, it is an open problem for machines to understand humor. In this paper, we build a new multimodal dataset for humor prediction that includes subtitles and video frames, as well as humor labels associated with video's timestamps. On top of it, we present a model to predict whether a subtitle causes laughter. Our model uses the visual modality through facial expression and character name recognition, together with the verbal modality, to explore how the visual modality helps. In addition, we use an attention mechanism to adjust the weight for each modality to facilitate humor prediction. Interestingly, our experimental results show that the performance boost by combinations of different modalities, and the attention mechanism and the model mostly relies on the verbal modality.

The availability of a large labelled dataset is a key requirement for applying deep learning methods to solve various computer vision tasks. In the context of understanding human activities, existing public datasets, while large in size, are often limited to a single RGB camera and provide only per-frame or per-clip action annotations. To enable richer analysis and understanding of human activities, we introduce IKEA ASM---a three million frame, multi-view, furniture assembly video dataset that includes depth, atomic actions, object segmentation, and human poses. Additionally, we benchmark prominent methods for video action recognition, object segmentation and human pose estimation tasks on this challenging dataset. The dataset enables the development of holistic methods, which integrate multi-modal and multi-view data to better perform on these tasks.

Pursuing a more coherent scene understanding towards real-time vision applications, single-stage instance segmentation has recently gained popularity, achieving a simpler and more efficient design than its two-stage counterparts. Besides, its global mask representation often leads to superior accuracy to the two-stage Mask R-CNN which has been dominant thus far. Despite the promising advances in single-stage methods, finer delineation of instance boundaries still remains unexcavated. Indeed, boundary information provides a strong shape representation that can operate in synergy with the fully-convolutional mask features of the single-stage segmented. In this work, we propose Boundary Basis based Instance Segmentation(B2Inst) to learn a global boundary representation that can complement existing global-mask-based methods that are often lacking high-frequency details. Besides, we devise a unified quality measure of both mask and boundary and introduce a network block that learns to score the per-instance predictions of itself. When applied to the strongest baselines in single-stage instance segmentation, our B2Inst leads to consistent improvements and accurately parse out the instance boundaries in a scene. Regardless of being single-stage or two-stage frameworks, we outperform the existing state-of-the-art methods on the COCO dataset with the same ResNet-50 and ResNet-101 backbones.

Localized Narratives is a dataset with detailed natural language descriptions of images paired with mouse traces that provide a sparse, fine-grained visual grounding for phrases. We propose TReCS, a sequential model that exploits this grounding to generate images. TReCS uses descriptions to retrieve segmentation masks and predict object labels aligned with mouse traces. These alignments are used to select and position masks to generate a fully covered segmentation canvas; the final image is produced by a segmentation-to-image generator using this canvas. This multi-step, retrieval-based approach outperforms existing direct text-to-image generation models on both automatic metrics and human evaluations: overall, its generated images are more photo-realistic and better match descriptions.

This paper presents a 3D human pose estimation system that uses a stereo pair of 360deg sensors to capture the complete scene from a single location. The approach combines the advantages of omnidirectional capture, the accuracy of multiple view 3D pose estimation and the portability of monocular acquisition. Joint monocular belief maps for joint locations are estimated from 360deg images and are used to fit a 3D skeleton to each frame. Temporal data association and smoothing is performed to produce accurate 3D pose estimates throughout the sequence. We evaluate our system on the Panoptic Studio dataset, as well as real 360deg video for tracking multiple people, demonstrating an average Mean Per Joint Position Error of 124.73mm with 30cm baseline cameras. We also demonstrate improved capabilities over perspective and 360deg multi-view systems when presented with limited camera views of the subject.

Estimating 3D hand pose directly from RGB images is challenging but has gained steady progress recently by training deep models with annotated 3D poses. However annotating 3D poses is difficult and as such only a few 3D hand pose datasets are available, all with limited sample sizes. In this study, we propose a new framework of training 3D pose estimation models from RGB images without using explicit 3D annotations, i.e., trained with only 2D information. Our framework is motivated by two observations: 1) Videos provide richer information for estimating 3D poses as opposed to static images; 2) Estimated 3D poses ought to be consistent whether the videos are viewed in the forward order or reverse order. We leverage these two observations to develop a self-supervised learning model called temporal-aware self-supervised network (TASSN).By enforcing temporal consistency constraints, TASSN learns 3D hand poses and meshes from videos with only 2D keypoint position annotations. Experiments show that our model achieves surprisingly good results, with 3D estimation accuracy on par with the state-of-the-art models trained with 3D annotations, highlighting the benefit of the temporal consistency in constraining 3D prediction models.

Training deep learning models for accurate spatiotemporal recognition of facial expressions in videos requires significant computational resources. For practical reasons, 3D Convolutional Neural Networks (3D CNNs) are usually trained with relatively short clips randomly extracted from videos. However, such uniform sampling is generally sub-optimal because equal importance is assigned to each temporal clip. In this paper, we present a strategy for efficient video-based training of 3D CNNs. It relies on softmax temporal pooling and a weighted sampling mechanism to select the most relevant training clips. The proposed softmax strategy provides several advantages - a reduced computational complexity due to efficient clip sampling, and an improved accuracy since temporal weighting focuses on more relevant clips during both training and inference. Experimental results obtained with the proposed method on several facial expression recognition benchmarks show the benefits of focusing on more informative clips in training videos. In particular, our approach improves performance and computational cost by reducing the impact of inaccurate trimming and coarse annotation of videos, and heterogeneous distribution of visual information across time.

Video generation models have become increasingly popular in the last few years, however the standard 2D architectures used today lack natural spatio-temporal modelling capabilities. In this paper, we present a network architecture for video generation that models spatio-temporal consistency without resorting to costly 3D architectures. The architecture facilitates information exchange between neighboring time points, which improves the temporal consistency of both the high level structure as well as the low-level details of the generated frames. The approach achieves state-of-the-art quantitative performance, as measured by the inception score on the UCF-101 dataset as well as better qualitative results. We also introduce a new quantitative measure (S3) that uses downstream tasks for evaluation. Moreover, we present a new multi-label dataset MaisToy, which enables us to evaluate the generalization of the model.

The current research focus on Content-Based Video Retrieval requires higher-level video representation describing the long-range semantic dependencies of relevant incidents, events, etc. However, existing methods commonly process the frames of a video as individual images or short clips, making the modeling of long-range semantic dependencies difficult. In this paper, we propose TCA (Temporal Context Aggregation for Video Retrieval), a video representation learning network that incorporates long-range temporal information between frame-level features using the self-attention mechanism for video retrieval. To train it on video retrieval datasets, we propose a supervised contrastive learning method that performs automatic hard negative mining and utilizes the memory bank mechanism to increase the capacity of negative samples. Extensive experiments are conducted on multiple video retrieval tasks, such as CC_WEB_VIDEO, FIVR-200K, and EVVE. The proposed method shows a significant performance advantage ( 17% mAP on FIVR-200K) over state-of-the-art methods with video-level features, and deliver competitive results with 22x faster inference time comparing with frame-level features.

Some tasks, such as surface normals or single-view depth estimation, require per-pixel ground truth that is difficult to obtain on real images but easy to obtain on synthetic. However, models learned on synthetic images often do not generalize well to real images due to the domain shift. Our key idea to improve domain adaptation is to introduce a separate anchor task (such as facial landmarks) whose annotations can be obtained at no cost or are already available on both synthetic and real datasets. To further leverage the implicit relationship between the anchor and main tasks, we apply our HeadFreeze technique that learns the cross-task guidance on the source domain with the final network layers, and use it on the target domain. We evaluate our methods on surface normal estimation on two pairs of datasets (indoor scenes and faces) with two kinds of anchor tasks (semantic segmentation and facial landmarks). We show that blindly applying domain adaptation or training the auxiliary task on only one domain may hurt performance, while using anchor tasks on both domains is better behaved. Our HeadFreeze technique outperforms competing approaches, reaching performance in facial images on par with a recently popular surface normal estimation method using shape from shading domain knowledge.

Many deep learning models, developed in recent years, reach higher ImageNet accuracy than ResNet50, with fewer or comparable FLOPs count. While FLOPs are often seen as a proxy for network efficiency, when measuring actual GPU training and inference throughput, vanilla ResNet50 is usually significantly faster than its recent competitors, offering better throughput-accuracy trade-off. In this work, we introduce a series of architecture modifications that aim to boost neural networks' accuracy, while retaining their GPU training and inference efficiency. We first demonstrate and discuss the bottlenecks induced by FLOPs oriented optimizations. We then suggest alternative designs that better utilize GPU structure and assets. Finally, we introduce a new family of GPU-dedicated models, called TResNet, which achieves better accuracy and efficiency than previous ConvNets. Using a TResNet model, with similar GPU throughput to ResNet50, we reach 80.8% top-1 accuracy on ImageNet. Our TResNet models also transfer well and achieve state-of-the-art accuracy on competitive single-label classification datasets such as Stanford Cars (96.0%), CIFAR-10 (99.0%), CIFAR-100 (91.5%) and Oxford-Flowers (99.1%). TResNet models also achieve state-of-the-art results on a multi-label classification task, and perform well on object detection.

Regular pavement inspection plays a significant role in road maintenance for safety assurance. Existing methods mainly address the tasks of crack detection and segmentation that are only tailored for long-thin crack disease. However, there are many other types of diseases with a wider variety of sizes and patterns that are also essential to segment in practice, bringing more challenges towards fine-grained pavement inspection. In this paper, our goal is not only to automatically segment cracks, but also to segment other complex pavement diseases as well as typical landmarks (markings, runway lights, etc.) and commonly seen water/oil stains in a single model. To this end, we propose a three-stream boundary-aware network (TB-Net). It consists of three streams fusing the low-level spatial and the high-level contextual representations as well as the detailed boundary information. Specifically, the spatial stream captures rich spatial features. The context stream, where an attention mechanism is utilized, models the contextual relationships over local features. The boundary stream learns detailed boundaries using a global-gated convolution to further refine the segmentation outputs. The network is trained using a dual-task loss in an end-to-end manner, and experiments on a newly collected fine-grained pavement disease dataset show the effectiveness of our TB-Net.

Critical obstacles in training classifiers to detect facial actions are the limited sizes of annotated video databases and the relatively low frequencies of occurrence of many actions. To address these problems, we propose an approach that makes use of facial expression generation. Our approach reconstructs the 3D shape of the face from each video frame, aligns the 3D mesh to a canonical view, and then trains a GAN-based network to synthesize novel images with facial action units of interest. To evaluate this approach, a deep neural network was trained on two separate datasets: One network was trained on video of synthesized facial expressions generated from FERA17; the other network was trained on unaltered video from the same database. Both networks used the same train and validation partitions and were tested on the test partition of actual video from FERA17. The network trained on synthesized facial expressions outperformed the one trained on actual facial expressions and surpassed current state-of-the-art approaches.

State-of-the-art self-supervised learning approaches for monocular depth estimation usually suffer from scale ambiguity. They do not generalize well when applied on distance estimation for complex projection models such as in fisheye and omnidirectional cameras. This paper introduces a novel multi-task learning strategy to improve self-supervised monocular distance estimation on fisheye and pinhole camera images. Our contribution to this work is threefold: Firstly, we introduce a novel distance estimation network architecture using a self-attention based encoder coupled with robust semantic feature guidance to the decoder that can be trained in a one-stage fashion. Secondly, we integrate a generalized robust loss function, which improves performance significantly while removing the need for hyperparameter tuning with the reprojection loss. Finally, we reduce the artifacts caused by dynamic objects violating static world assumptions using a semantic masking strategy. We significantly improve upon the RMSE of previous work on fisheye by a 25% reduction in RMSE. As there is little work on fisheye cameras, we evaluated the proposed method on KITTI using a pinhole model. We achieved state-of-the-art performance among self-supervised methods without requiring an external scale estimation.

A significant challenge in video understanding is posed by the high dimensionality of the input, which induces large computational cost and high memory footprints. Deep convolutional models operating on video apply pooling and striding to reduce feature dimensionality and to increase the receptive field. However, despite these strategies, modern approaches cannot effectively leverage spatiotemporal structure over long temporal extents. In this paper we introduce an approach that reduces a video of 10 seconds to a sparse graph of only 160 feature nodes such that efficient inference in this graph produces state-of-the-art accuracy on challenging action recognition datasets. The nodes of our graph are semantic supervoxels that capture the spatiotemporal structure of objects and motion cues in the video, while edges between nodes encode spatiotemporal relations and feature similarity. We demonstrate that a shallow network that interleaves graph convolution and graph pooling on this compact representation implements an effective mechanism of relational reasoning yielding strong recognition results on both Charades and Something-Something.

Localization and reconstruction of underground targets, the problem of estimating the position and geometry of the objects from Ground Penetration Radar (GPR), still lies at the core of non-destructive testing (NDT). In this paper, we present MigrationNet, a learning-based approach to detect and visualize subsurface objects. Compared with the existing learning-based method of GPR, our proposed approach could not only detect the hyperbola feature in the raw B-scan image but also interpret hyperbola features into the cross-section image of subsurface pipes. Furthermore, to compare the proposed method with the conventional back-projection methods for GPR data interpretation, a synthetic GPR dataset that mimics the real NDT environment is also introduced in this work. The study indicates the effectiveness of our method, it uses less GPR data for underground pipes reconstruction, produces better GPR imaging results with less computation, and shows the robustness to noise.

We propose a method for synthesizing eye images from segmentation masks with a desired style. The style encompasses attributes such as skin color, texture, iris color, and personal identity. Our approach generates an eye image that is consistent with a given segmentation mask and has the attributes of the input style image. We apply our method to data augmentation as well as to gaze redirection. The previous techniques of synthesizing real eye images from synthetic eye images for data augmentation lacked control over the generated attributes. We demonstrate the effectiveness of the proposed method in synthesizing realistic eye images with given characteristics corresponding to the synthetic labels for data augmentation, which is further useful for various tasks such as gaze estimation, eye image segmentation, pupil detection, etc. We also show how our approach can be applied to gaze redirection using only synthetic gaze labels, improving the previous state of the art results. The main contributions of our paper are i) a novel approach for Style-Based eye image generation from segmentation mask; ii) the use of this approach for gaze-redirection without the need for gaze annotated real eye images

Existing Image Captioning (IC) systems model words as atomic units in captions and are unable to exploit the structural information in the words. This makes representation of rare words very difficult and out-of-vocabulary words impossible. Moreover, to avoid computational complexity, existing IC models operate over a modest sized vocabulary of frequent words, such that the identity of rare words is lost. In this work we address this common limitation of IC systems in dealing with rare words in the corpora. We decompose words into smaller constituent units `subwords' and represent captions as a sequence of subwords instead of words. This helps represent all words in the corpora using a significantly lower subword vocabulary, leading to better parameter learning. Using subword language modeling, our captioning system improves various metric scores, with a training vocabulary size approximately 90% less than the baseline and various state-of-the-art word-level models. Our quantitative and qualitative results and analysis signify the efficacy of our proposed approach.