ICCV 2021 Papers

We present a domain- and user-preference-agnostic approach to detect highlightable excerpts from human-centric videos. Our method works on the graph-based representation of multiple observable human-centric modalities in the videos, such as poses and faces. We use an autoencoder network equipped with spatial-temporal graph convolutions to detect human activities and interactions based on these modalities. We train our network to map the activity- and interaction-based latent structural representations of the different modalities to per-frame highlight scores based on the representativeness of the frames. We use these scores to compute which frames to highlight and stitch contiguous frames to produce the excerpts. We train our network on the large-scale AVA-Kinetics action dataset and evaluate it on four benchmark video highlight datasets: DSH, TVSum, PHD^2, and SumMe. We observe a 4-12% improvement in the mean average precision of matching the human-annotated highlights over state-of-the-art methods in these datasets, without requiring any user-provided preferences or dataset-specific fine-tuning.

Optical flow is inherently a 2D search problem, and thus the computational complexity grows quadratically with respect to the search window, making large displacements matching infeasible for high-resolution images. In this paper, we take inspiration from Transformers and propose a new method for high-resolution optical flow estimation with significantly less computation. Specifically, a 1D attention operation is first applied in the vertical direction of the target image, and then a simple 1D correlation in the horizontal direction of the attended image is able to achieve 2D correspondence modeling effect. The directions of attention and correlation can also be exchanged, resulting in two 3D cost volumes that are concatenated for optical flow estimation. The novel 1D formulation empowers our method to scale to very high-resolution input images while maintaining competitive performance. Extensive experiments on Sintel, KITTI and real-world 4K (2160 x 3840) resolution images demonstrated the effectiveness and superiority of our proposed method. Code and models are available at https://github.com/haofeixu/flow1d.

End-to-end discriminative trackers improve the state of the art significantly, yet the improvement in robustness and efficiency is restricted by the conventional discriminative model, i.e., least-squares based regression. In this paper, we present DTT, a novel single-object discriminative tracker, based on an encoder-decoder Transformer architecture. By self- and encoder-decoder attention mechanisms, our approach is able to exploit the rich scene information in an end-to-end manner, effectively removing the need for hand-designed discriminative models. In online tracking, given a new test frame, dense prediction is performed at all spatial positions. Not only location, but also bounding box of the target object is obtained in a robust fashion, streamlining the discriminative tracking pipeline. DTT is conceptually simple and easy to implement. It yields state-of-the-art performance on four popular benchmarks including GOT-10k, LaSOT, NfS, and TrackingNet while running at over 50 FPS, confirming its effectiveness and efficiency. We hope DTT may provide a new perspective for single-object visual tracking.

Image completion has made tremendous progress with convolutional neural networks (CNNs), because of their powerful texture modeling capacity. However, due to some inherent properties (eg, local inductive prior, spatial-invariant kernels), CNNs do not perform well in understanding global structures or naturally support pluralistic completion. Recently, transformers demonstrate their power in modeling the long-term relationship and generating diverse results, but their computation complexity is quadratic to input length, thus hampering the application in processing high-resolution images. This paper brings the best of both worlds to pluralistic image completion: appearance prior reconstruction with transformer and texture replenishment with CNN. The former transformer recovers pluralistic coherent structures together with some coarse textures, while the latter CNN enhances the local texture details of coarse priors guided by the high-resolution masked images. The proposed method vastly outperforms state-of-the-art methods in terms of three aspects: 1) large performance boost on image fidelity even compared to deterministic completion methods; 2) better diversity and higher fidelity for pluralistic completion; 3) exceptional generalization ability on large masks and generic dataset, like ImageNet. Code and pre-trained models have been publicly released at https://github.com/raywzy/ICT.

A high quality disparity remapping method that preserves 2D shapes and 3D structures, and adjusts disparities of important objects in stereo image pairs is proposed. It is formulated as a constrained optimization problem, whose solution is challenging, since we need to meet multiple requirements of disparity remapping simultaneously. The one-stage optimization process either degrades the quality of important objects or introduces serious distortions in background regions. To address this challenge, we propose a two-stage warping process to solve it. In the first stage, we develop a warping model that finds the optimal warping grids for important objects to fulfill multiple requirements of disparity remapping. In the second stage, we derive another warping model to refine warping results in less important regions by eliminating serious distortions in shape, disparity and 3D structure. The superior performance of the proposed method is demonstrated by experimental results

The goal of object navigation is to reach the expected objects according to visual information in the unseen environments. Previous works usually implement deep models to train an agent to predict actions in real-time. However, in the unseen environment, when the target object is not in egocentric view, the agent may not be able to make wise decisions due to the lack of guidance. In this paper, we propose a hierarchical object-to-zone (HOZ) graph to guide the agent in a coarse-to-fine manner, and an online-learning mechanism is also proposed to update HOZ according to the real-time observation in new environments. In particular, the HOZ graph is composed of scene nodes, zone nodes and object nodes. With the pre-learned HOZ graph, the real-time observation and the target goal, the agent can constantly plan an optimal path from zone to zone. In the estimated path, the next potential zone is regarded as sub-goal, which is also fed into the deep reinforcement learning model for action prediction. Our methods are evaluated on the AI2-Thor simulator. In addition to widely used evaluation metrics SR and SPL, we also propose a new evaluation metric of SAE that focuses on the effective action rate. Experimental results demonstrate the effectiveness and efficiency of our proposed method. The code is available at https://github.com/sx-zhang/HOZ.git.

We present Hierarchical Memory Matching Network (HMMN) for semi-supervised video object segmentation. Based on a recent memory-based method [33], we propose two advanced memory read modules that enable us to perform memory reading in multiple scales while exploiting temporal smoothness. We first propose a kernel guided memory matching module that replaces the non-local dense memory read, commonly adopted in previous memory-based methods. The module imposes the temporal smoothness constraint in the memory read, leading to accurate memory retrieval. More importantly, we introduce a hierarchical memory matching scheme and propose a top-k guided memory matching module in which memory read on a fine-scale is guided by that on a coarse-scale. With the module, we perform memory read in multiple scales efficiently and leverage both high-level semantic and low-level fine-grained memory features to predict detailed object masks. Our network achieves state-of-the-art performance on the validation sets of DAVIS 2016/2017 (90.8% and 84.7%) and YouTube-VOS 2018/2019 (82.6% and 82.5%), and test-dev set of DAVIS 2017 (78.6%). The source code and model are available online: https://github.com/Hongje/HMMN.

This paper addresses the problem of 3D human body shape and pose estimation from an RGB image. This is often an ill-posed problem, since multiple plausible 3D bodies may match the visual evidence present in the input - particularly when the subject is occluded. Thus, it is desirable to estimate a distribution over 3D body shape and pose conditioned on the input image instead of a single 3D reconstruction. We train a deep neural network to estimate a hierarchical matrix-Fisher distribution over relative 3D joint rotation matrices (i.e. body pose), which exploits the human body's kinematic tree structure, as well as a Gaussian distribution over SMPL body shape parameters. To further ensure that the predicted shape and pose distributions match the visual evidence in the input image, we implement a differentiable rejection sampler to impose a reprojection loss between ground-truth 2D joint coordinates and samples from the predicted distributions, projected onto the image plane. We show that our method is competitive with the state-of-the-art in terms of 3D shape and pose metrics on the SSP-3D and 3DPW datasets, while also yielding a structured probability distribution over 3D body shape and pose, with which we can meaningfully quantify prediction uncertainty and sample multiple plausible 3D reconstructions to explain a given input image.

Deep learning has made tremendous success in computer vision, natural language processing and even visual-semantic learning, which requires a huge amount of labeled training data. Nevertheless, the goal of human-level intelligence is to enable a model to quickly obtain an in-depth understanding given a small number of samples, especially with heterogeneity in the multi-modal scenarios such as visual question answering and image captioning. In this paper, we study the few-shot visual-semantic learning and present the Hierarchical Graph ATtention network (HGAT). This two-stage network models the intra- and inter-modal relationships with limited image-text samples. The main contributions of HGAT can be summarized as follows: 1) it sheds light on tackling few-shot multi-modal learning problems, which focuses primarily, but not exclusively on visual and semantic modalities, through better exploitation of the intra-relationship of each modality and an attention-based co-learning framework between modalities using a hierarchical graph-based architecture; 2) it achieves superior performance on both visual question answering and image captioning in the few-shot setting; 3) it can be easily extended to the semi-supervised setting where image-text samples are partially unlabeled. We show via extensive experiments that HGAT delivers state-of-the-art performance on three widely-used benchmarks of two visual-semantic learning tasks.

Data-driven sky models have gained much attention in outdoor illumination prediction recently, showing superior performance against analytical models. However, naively compressing an outdoor panorama into a low-dimensional latent vector, as existing models have done, causes two major problems. One is the mutual interference between the HDR intensity of the sun and the complex textures of the surrounding sky, and the other is the lack of fine-grained control over independent lighting factors due to the entangled representation. To address these issues, we propose a hierarchical disentangled sky model (HDSky) for outdoor illumination prediction. With this model, any outdoor panorama can be hierarchically disentangled into several factors based on three well-designed autoencoders. The first autoencoder compresses each sunny panorama into a sky vector and a sun vector with some constraints. The second autoencoder and the third autoencoder further disentangle the sun intensity and the sky intensity from the sun vector and the sky vector with several customized loss functions respectively. Moreover, a unified framework is designed to predict all-weather sky information from a single outdoor image. Through extensive experiments, we demonstrate that the proposed model significantly improves the accuracy of outdoor illumination prediction. It also allows users to intuitively edit the predicted panorama (e.g., changing the position of the sun while preserving others), without sacrificing physical plausibility.

Normalizing flows have recently demonstrated promising results for low-level vision tasks. For image super-resolution (SR), it learns to predict diverse photo-realistic high-resolution (HR) images from the low-resolution (LR) image rather than learning a deterministic mapping. For image rescaling, it achieves high accuracy by jointly modelling the downscaling and upscaling processes. While existing approaches employ specialized techniques for these two tasks, we set out to unify them in a single formulation. In this paper, we propose the hierarchical conditional flow (HCFlow) as a unified framework for image SR and image rescaling. More specifically, HCFlow learns a bijective mapping between HR and LR image pairs by modelling the distribution of the LR image and the rest high-frequency component simultaneously. In particular, the high-frequency component is conditional on the LR image in a hierarchical manner. To further enhance the performance, other losses such as perceptual loss and GAN loss are combined with the commonly used negative log-likelihood loss in training. Extensive experiments on general image SR, face image SR and image rescaling have demonstrated that the proposed HCFlow achieves state-of-the-art performance in terms of both quantitative metrics and visual quality.

Instance segmentation on point clouds is a fundamental task in 3D scene perception. In this work, we propose a concise clustering-based framework named HAIS, which makes full use of spatial relation of points and point sets. Considering clustering-based methods may result in over-segmentation or under-segmentation, we introduce the hierarchical aggregation to progressively generate instance proposals, i.e., point aggregation for preliminarily clustering points to sets and set aggregation for generating complete instances from sets. Once the complete 3D instances are obtained, a sub-network of intra-instance prediction is adopted for noisy points filtering and mask quality scoring. HAIS is fast (only 410ms per frame on Titan X) and does not require non-maximum suppression. It ranks 1st on the ScanNet v2 benchmark, achieving the highest 69.9% AP50 and surpassing previous state-of-the-art (SOTA) methods by a large margin. Besides, the SOTA results on the S3DIS dataset validate the good generalization ability. Code is available at https://github.com/hustvl/HAIS.

Video-Text Retrieval has been a hot research topic with the growth of multimedia data on the internet. Transformer for video-text learning has attracted increasing attention due to its promising performance. However, existing cross-modal transformer approaches typically suffer from two major limitations: 1) Exploitation of the transformer architecture where different layers have different feature characteristics is limited; 2) End-to-end training mechanism limits negative sample interactions in a mini-batch. In this paper, we propose a novel approach named Hierarchical Transformer (HiT) for video-text retrieval. HiT performs Hierarchical Cross-modal Contrastive Matching in both feature-level and semantic-level, achieving multi-view and comprehensive retrieval results. Moreover, inspired by MoCo, we propose Momentum Cross-modal Contrast for cross-modal learning to enable large-scale negative sample interactions on-the-fly, which contributes to the generation of more precise and discriminative representations. Experimental results on the three major Video-Text Retrieval benchmark datasets demonstrate the advantages of our method.

Image hiding aims to hide a secret image into a cover image in an imperceptible way, and then recover the secret image perfectly at the receiver end. Capacity, invisibility and security are three primary challenges in image hiding task. This paper proposes a novel invertible neural network (INN) based framework, HiNet, to simultaneously overcome the three challenges in image hiding. For large capacity, we propose an inverse learning mechanism by simultaneously learning the image concealing and revealing processes. Our method is able to achieve the concealing of a full-size secret image into a cover image with the same size. For high invisibility, instead of pixel domain hiding, we propose to hide the secret information in wavelet domain. Furthermore, we propose a new low-frequency wavelet loss to constrain that secret information is hidden in high-frequency wavelet sub-bands, which significantly improves the hiding security. Experimental results show that our HiNet significantly outperforms other state-of-the-art image hiding methods, with more than 10 dB PSNR improvement in secret image recovery on ImageNet, COCO and DIV2K datasets. Codes are available at https://github.com/TomTomTommi/HiNet.

Most existing Siamese-based tracking methods execute the classification and regression of the target object based on the similarity maps. However, they either employ a single map from the last convolutional layer which degrades the localization accuracy in complex scenarios or separately use multiple maps for decision making, introducing intractable computations for aerial mobile platforms. Thus, in this work, we propose an efficient and effective hierarchical feature transformer (HiFT) for aerial tracking. Hierarchical similarity maps generated by multi-level convolutional layers are fed into the feature transformer to achieve the interactive fusion of spatial (shallow layers) and semantics cues (deep layers). Consequently, not only the global contextual information can be raised, facilitating the target search, but also our end-to-end architecture with the transformer can efficiently learn the interdependencies among multi-level features, thereby discovering a tracking-tailored feature space with strong discriminability. Comprehensive evaluations on four aerial benchmarks have proven the effectiveness of HiFT. Real-world tests on the aerial platform have strongly validated its practicability with a real-time speed. Our code is available at https://github.com/vision4robotics/HiFT.

The crucial problem in vehicle re-identification is to find the same vehicle identity when reviewing this object from cross-view cameras, which sets a higher demand for learning viewpoint-invariant representations. In this paper, we propose to solve this problem from two aspects: constructing robust feature representations and proposing camera-sensitive evaluations. We first propose a novel Heterogeneous Relational Complement Network (HRCN) by incorporating region-specific features and cross-level features as complements for the original high-level output. Considering the distributional differences and semantic misalignment, we propose graph-based relation modules to embed these heterogeneous features into one unified high-dimensional space. On the other hand, considering the deficiencies of cross-camera evaluations in existing measures (i.e., CMC and AP), we then propose a Cross-camera Generalization Measure (CGM) to improve the evaluations by introducing position-sensitivity and cross-camera generalization penalties. We further construct a new benchmark of existing models with our proposed CGM and experimental results reveal that our proposed HRCN model achieves new state-of-the-art in VeRi-776, VehicleID, and VERI-Wild.

Recent attempts to solve the problem of head reenactment using a single reference image have shown promising results. However, most of them either perform poorly in terms of photo-realism, or fail to meet the identity preservation problem, or do not fully transfer the driving pose and expression. We propose HeadGAN, a novel system that conditions synthesis on 3D face representations, which can be extracted from any driving video and adapted to the facial geometry of any reference image, disentangling identity from expression. We further improve mouth movements, by utilising audio features as a complementary input. The 3D face representation enables HeadGAN to be further used as an efficient method for compression and reconstruction and a tool for expression and pose editing.

Existing methods for multi-modal domain translation learn to embed the input images into a domain-invariant "content" space and a domain-specific "style" space from which novel images can be synthesized. Rather than learning to embed the RGB image from scratch we propose deriving our content representation from conditioning data produced by pretrained off-the-shelf networks. Motivated by the inherent ambiguity of "content", which has different meanings depending on the desired level of abstraction, this approach gives intuitive control over which aspects of content are preserved across domains. We evaluate our method on traditional, well-aligned, datasets such as CelebA-HQ, and propose two novel datasets for evaluation on more complex scenes: ClassicTV and FFHQ-WildCrops. Our approach, which we call Sensorium, enables higher quality domain translation for complex scenes than prior work.

We propose a novel transformer-based styled handwritten text image generation approach, HWT, that strives to learn both style-content entanglement as well as global and local style patterns. The proposed HWT captures the long and short range relationships within the style examples through a self-attention mechanism, thereby encoding both global and local style patterns. Further, the proposed transformer-based HWT comprises an encoder-decoder attention that enables style-content entanglement by gathering the style features of each query character. To the best of our knowledge, we are the first to introduce a transformer-based network for styled handwritten text generation. Our proposed HWT generates realistic styled handwritten text images and outperforms the state-of-the-art demonstrated through extensive qualitative, quantitative and human-based evaluations. The proposed HWT can handle arbitrary length of text and any desired writing style in a few-shot setting. Further, our HWT generalizes well to the challenging scenario where both words and writing style are unseen during training, generating realistic styled handwritten text images.