CVPR 2022 Papers

3D room layout estimation by a single panorama using deep neural networks has made great progress. However, previous approaches can not obtain efficient geometry awareness of room layout with the only latitude of boundaries or horizon-depth. We present that using horizon-depth along with room height can obtain omnidirectional-geometry awareness of room layout in both horizontal and vertical directions. In addition, we propose a planar-geometry aware loss function with normals and gradients of normals to supervise the planeness of walls and turning of corners. We propose an efficient network, LGT-Net, for room layout estimation, which contains a novel Transformer architecture called SWG-Transformer to model geometry relations. SWG-Transformer consists of (Shifted) Window Blocks and Global Blocks to combine the local and global geometry relations. Moreover, we design a novel relative position embedding of Transformer to enhance the spatial identification ability for the panorama. Experiments show that the proposed LGT-Net achieves better performance than current state-of-the-arts (SOTA) on benchmark datasets.

Gesture recognition plays an important role in natural human-computer interaction and sign language recognition. Existing research on gesture recognition is limited to close-range interaction such as vehicle gesture control and face-to-face communication. To apply gesture recognition to long-distance interactive scenes such as meetings and smart homes, a large RGB-D video dataset LD-ConGR is established in this paper. LD-ConGR is distinguished from existing gesture datasets by its long-distance gesture collection, fine-grained annotations, and high video quality. Specifically, 1) the farthest gesture provided by the LD-ConGR is captured 4m away from the camera while existing gesture datasets collect gestures within 1m from the camera; 2) besides the gesture category, the temporal segmentation of gestures and hand location are also annotated in LD-ConGR; 3) videos are captured at high resolution (1280x720 for color streams and 640x576 for depth streams) and high frame rate (30 fps). On top of the LD-ConGR, a series of experimental and studies are conducted, and the proposed gesture region estimation and key frame sampling strategies are demonstrated to be effective in dealing with long-distance gesture recognition and the uncertainty of gesture duration. The dataset and experimental results presented in this paper are expected to boost the research of long-distance gesture recognition. The dataset is available at https://github.com/Diananini/LD-ConGR-CVPR2022.

Recent learning-based lossless image compression methods encode an image in the unit of subimages and achieve comparable performances to conventional non-learning algorithms. However, these methods do not consider the performance drop in the high-frequency region, giving equal consideration to the low and high-frequency areas. In this paper, we propose a new lossless image compression method that proceeds the encoding in a coarse-to-fine manner to separate and process low and high-frequency regions differently. We initially compress the low-frequency components and then use them as additional input for encoding the remaining high-frequency region. The low-frequency components act as a strong prior in this case, which leads to improved estimation in the high-frequency area. In addition, we design the frequency decomposition process to be adaptive to color channel, spatial location, and image characteristics. As a result, our method derives an image-specific optimal ratio of low/high-frequency components. Experiments show that the proposed method achieves state-of-the-art performance for benchmark high-resolution datasets.

Referring image segmentation is a fundamental vision-language task that aims to segment out an object referred to by a natural language expression from an image. One of the key challenges behind this task is leveraging the referring expression for highlighting relevant positions in the image. A paradigm for tackling this problem is to leverage a powerful vision-language ("cross-modal") decoder to fuse features independently extracted from a vision encoder and a language encoder. Recent methods have made remarkable advancements in this paradigm by exploiting Transformers as cross-modal decoders, concurrent to the Transformer's overwhelming success in many other vision-language tasks. Adopting a different approach in this work, we show that significantly better cross-modal alignments can be achieved through the early fusion of linguistic and visual features in intermediate layers of a vision Transformer encoder network. By conducting cross-modal feature fusion in the visual feature encoding stage, we can leverage the well-proven correlation modeling power of a Transformer encoder for excavating helpful multi-modal context. This way, accurate segmentation results are readily harvested with a light-weight mask predictor. Without bells and whistles, our method surpasses the previous state-of-the-art methods on RefCOCO, RefCOCO+, and G-Ref by large margins.

We present LASER, an image-based Monte Carlo Localization (MCL) framework for 2D floor maps. LASER introduces the concept of latent space rendering, where 2D pose hypotheses on the floor map are directly rendered into a geometrically-structured latent space by aggregating viewing ray features. Through a tightly coupled rendering codebook scheme, the viewing ray features are dynamically determined at rendering-time based on their geometries (i.e. length, incident-angle), endowing our representation with view-dependent fine-grain variability. Our codebook scheme effectively disentangles feature encoding from rendering, allowing the latent space rendering to run at speeds above 10KHz. Moreover, through metric learning, our geometrically-structured latent space is common to both pose hypotheses and query images with arbitrary field of views. As a result, LASER achieves state-of-the-art performance on large-scale indoor localization datasets (i.e. ZInD and Structured3D) for both panorama and perspective image queries, while significantly outperforming existing learning-based methods in speed.

Adversarial training (AT) is always formulated as a minimax problem, of which the performance depends on the inner optimization that involves the generation of adversarial examples (AEs). Most previous methods adopt Projected Gradient Decent (PGD) with manually specifying attack parameters for AE generation. A combination of the attack parameters can be referred to as an attack strategy. Several works have revealed that using a fixed attack strategy to generate AEs during the whole training phase limits the model robustness and propose to exploit different attack strategies at different training stages to improve robustness. But those multi-stage hand-crafted attack strategies need much domain expertise, and the robustness improvement is limited. In this paper, we propose a novel framework for adversarial training by introducing the concept of "learnable attack strategy", dubbed LAS-AT, which learns to automatically produce attack strategies to improve the model robustness. Our framework is composed of a target network that uses AEs for training to improve robustness, and a strategy network that produces attack strategies to control the AE generation. Experimental evaluations on three benchmark databases demonstrate the superiority of the proposed method, and the proposed method outperforms state-of-the-art adversarial training methods.

We propose a novel method for solving regression tasks using few-shot or weak supervision. At the core of our method is the fundamental observation that GANs are incredibly successful at encoding semantic information within their latent space, even in a completely unsupervised setting. For modern generative frameworks, this semantic encoding manifests as smooth, linear directions which affect image attributes in a disentangled manner. These directions have been widely used in GAN-based image editing. In this work, we leverage them for few-shot regression. Specifically, we make the simple observation that distances traversed along such directions are good features for downstream tasks - reliably gauging the magnitude of a property in an image. In the absence of explicit supervision, we use these distances to solve tasks such as sorting a collection of images, and ordinal regression. With a few labels -- as little as two -- we calibrate these distances to real-world values and convert a pre-trained GAN into a state-of-the-art few-shot regression model. This enables solving regression tasks on datasets and attributes which are difficult to produce quality supervision for. Extensive experimental evaluations demonstrate that our method can be applied across a wide range of domains, leverage multiple latent direction discovery frameworks, and achieve state-of-the-art results in few-shot and low-supervision settings, even when compared to methods designed to tackle a single task. Code is available on our project website.

Previous super-resolution (SR) approaches often formulate SR as a regression problem and pixel wise restoration, which leads to a blurry and unreal SR output. Recent works combine adversarial loss with pixel-wise loss to train a GAN-based model or introduce normalizing flows into SR problems to generate more realistic images. As another powerful generative approach, autoregressive (AR) model has not been noticed in low level tasks due to its limitation. Based on the fact that given the structural information, the textural details in the natural images are locally related without long term dependency, in this paper we propose a novel autoregressive model-based SR approach, namely LAR-SR, which can efficiently generate realistic SR images using a novel local autoregressive (LAR) module. The proposed LAR module can sample all the patches of textural components in parallel, which greatly reduces the time consumption. In addition to high time efficiency, it is also able to leverage contextual information of pixels and can be optimized with a consistent loss. Experimental results on the widely-used datasets show that the proposed LAR-SR approach achieves superior performance on the visual quality and quantitative metrics compared with other generative models such as GAN, Flow, and is competitive with the mixture generative model.

Point cloud completion aims at completing geometric and topological shapes from a partial observation. However, some topology of the original shape is missing, existing methods directly predict the location of complete points, without predicting structured and topological information of the complete shape, which leads to inferior performance. To better tackle the missing topology part, we propose LAKe-Net, a novel topology-aware point cloud completion model by localizing aligned keypoints, with a novel Keypoints-Skeleton-Shape prediction manner. Specifically, our method completes missing topology using three steps: Aligned Keypoint Localization. An asymmetric keypoint locator, including an unsupervised multi-scale keypoint detector and a complete keypoint generator, is proposed for localizing aligned keypoints from complete and partial point clouds. We theoretically prove that the detector can capture aligned keypoints for objects within a sub-category. Surface-skeleton Generation. A new type of skeleton, named Surface-skeleton, is generated from keypoints based on geometric priors to fully represent the topological information captured from keypoints and better recover the local details. Shape Refinement. We design a refinement subnet where multi-scale surface-skeletons are fed into each recursive skeleton-assisted refinement module to assist the completion process. Experimental results show that our method achieves the state-of-the-art performance on point cloud completion.

Mining precise class-aware attention maps, a.k.a, class activation maps, is essential for weakly supervised semantic segmentation. In this paper, we present L2G, a simple online local-to-global knowledge transfer framework for high-quality object attention mining. We observe that classification models can discover object regions with more details when replacing the input image with its local patches. Taking this into account, we first leverage a local classification network to extract attentions from multiple local patches randomly cropped from the input image. Then, we utilize a global network to learn complementary attention knowledge across multiple local attention maps online. Our framework conducts the global network to learn the captured rich object detail knowledge from a global view and thereby produces high-quality attention maps that can be directly used as pseudo annotations for semantic segmentation networks. Experiments show that our method attains 72.1% and 44.2% mIoU scores on the validation set of PASCAL VOC 2012 and MS COCO 2014, respectively, setting new state-of-the-art records. Code is available at https://github.com/PengtaoJiang/L2G.

Far beyond learning long-range interactions of natural language, transformers are becoming the de-facto standard for many vision tasks with their power and scalability. Especially with cross-modal tasks between image and text, vector quantized variational autoencoders (VQ-VAEs) are widely used to make a raw RGB image into a sequence of feature vectors. To better leverage the correlation between image and text, we propose L-Verse, a novel architecture consisting of feature-augmented variational autoencoder (AugVAE) and bidirectional auto-regressive transformer (BiART) for image-to-text and text-to-image generation. Our AugVAE shows the state-of-the-art reconstruction performance on ImageNet1K validation set, along with the robustness to unseen images in the wild. Unlike other models, BiART can distinguish between image (or text) as a conditional reference and a generation target. L-Verse can be directly used for image-to-text or text-to-image generation without any finetuning or extra object detection framework. In quantitative and qualitative experiments, L-Verse shows impressive results against previous methods in both image-to-text and text-to-image generation on MS-COCO Captions. We furthermore assess the scalability of L-Verse architecture on Conceptual Captions and present the initial result of bidirectional vision-language representation learning on general domain.

Data is the driving force of machine learning, with the amount and quality of training data often being more important for the performance of a system than architecture and training details. But collecting, processing and annotating real data at scale is difficult, expensive, and frequently raises additional privacy, fairness and legal concerns. Synthetic data is a powerful tool with the potential to address these shortcomings: 1) it is cheap 2) supports rich ground-truth annotations 3) offers full control over data and 4) can circumvent or mitigate problems regarding bias, privacy and licensing. Unfortunately, software tools for effective data generation are less mature than those for architecture design and training, which leads to fragmented generation efforts. To address these problems we introduce Kubric, an open-source Python framework that interfaces with PyBullet and Blender to generate photo-realistic scenes, with rich annotations, and seamlessly scales to large jobs distributed over thousands of machines, and generating TBs of data. We demonstrate the effectiveness of Kubric by presenting a series of 11 different generated datasets for tasks ranging from studying 3D NeRF models to optical flow estimation. We release Kubric, the used assets, all of the generation code, as well as the rendered datasets for reuse and modification.

Facial action unit (AU) recognition is formulated as a supervised learning problem by recent works. However, the complex labeling process makes it challenging to provide AU annotations for large amounts of facial images. To remedy this, we utilize AU labeling rules defined by the Facial Action Coding System (FACS) to design a novel knowledge-driven self-supervised representation learning framework for AU recognition. The representation encoder is trained using large amounts of facial images without AU annotations. AU labeling rules are summarized from FACS to design facial partition manners and determine correlations between facial regions. The method utilizes a backbone network to extract local facial area representations and a project head to map the representations into a low-dimensional latent space. In the latent space, a contrastive learning component leverages the inter-area difference to learn AU-related local representations while maintaining intra-area instance discrimination. Correlations between facial regions summarized from AU labeling rules are also explored to further learn representations using a predicting learning component. Evaluation on two benchmark databases demonstrates that the learned representation is powerful and data-efficient for AU recognition.

Recently, the semantics of scene text has been proven to be essential in fine-grained image classification. However, the existing methods mainly exploit the literal meaning of scene text for fine-grained recognition, which might be irrelevant when it is not significantly related to objects/scenes. We propose an end-to-end trainable network that mines implicit contextual knowledge behind scene text image and enhance the semantics and correlation to fine-tune the image representation. Unlike the existing methods, our model integrates three modalities: visual feature extraction, text semantics extraction, and correlating background knowledge to fine-grained image classification. Specifically, we employ KnowBert to retrieve relevant knowledge for semantic representation and combine it with image features for fine-grained classification. Experiments on two benchmark datasets, Con-Text, and Drink Bottle, show that our method outperforms the state-of-the-art by 3.72% mAP and 5.39% mAP, respectively. To further validate the effectiveness of the proposed method, we create a new dataset on crowd activity recognition for the evaluation. The source code, new dataset, and pre-trained models of this work will be publicly available.

There is a growing discrepancy in computer vision between large-scale models that achieve state-of-the-art performance and models that are affordable in practical applications. In this paper we address this issue and significantly bridge the gap between these two types of models. Throughout our empirical investigation we do not aim to necessarily propose a new method, but strive to identify a robust and effective recipe for making state-of-the-art large scale models affordable in practice. We demonstrate that, when performed correctly, knowledge distillation can be a powerful tool for reducing the size of large models without compromising their performance. In particular, we uncover that there are certain implicit design choices, which may drastically affect the effectiveness of distillation. Our key contribution is the explicit identification of these design choices, which were not previously articulated in the literature. We back up our findings by a comprehensive empirical study, demonstrate compelling results on a wide range of vision datasets and, in particular, obtain a state-of-the-art ResNet-50 model for ImageNet, which achieves 82.8% top-1 accuracy.

Knowledge distillation becomes a de facto standard to improve the performance of small neural networks. Most of the previous works propose to regress the representational features from the teacher to the student in a one-to-one spatial matching fashion. However, people tend to overlook the fact that, due to the architecture differences, the semantic information on the same spatial location usually vary. This greatly undermines the underlying assumption of the one-to-one distillation approach. To this end, we propose a novel one-to-all spatial matching knowledge distillation approach. Specifically, we allow each pixel of the teacher feature to be distilled to all spatial locations of the student features given its similarity, which is generated from a target-aware transformer. Our approach surpasses the state-of-the-art methods by a significant margin on various computer vision benchmarks, such as ImageNet, Pascal VOC and COCOStuff10k. Code is available at https://github.com/sihaoevery/TaT.

Knowledge distillation aims to compress a powerful yet cumbersome teacher model into a lightweight student model without much sacrifice of performance. For this purpose, various approaches have been proposed over the past few years, generally with elaborately designed knowledge representations, which in turn increase the difficulty of model development and interpretation. In contrast, we empirically show that a simple knowledge distillation technique is enough to significantly narrow down the teacher-student performance gap. We directly reuse the discriminative classifier from the pre-trained teacher model for student inference and train a student encoder through feature alignment with a single L2 loss. In this way, the student model is able to achieve exactly the same performance as the teacher model provided that their extracted features are perfectly aligned. An additional projector is developed to help the student encoder match with the teacher classifier, which renders our technique applicable to various teacher and student architectures. Extensive experiments demonstrate that our technique achieves state-of-the-art results at the modest cost of compression ratio due to the added projector.

Large-scale pre-training has been proven to be crucial for various computer vision tasks. However, with the increase of pre-training data amount, model architecture amount, and the private/inaccessible data, it is not very efficient or possible to pre-train all the model architectures on large-scale datasets. In this work, we investigate an alternative strategy for pre-training, namely Knowledge Distillation as Efficient Pre-training (KDEP), aiming to efficiently transfer the learned feature representation from existing pre-trained models to new student models for future downstream tasks. We observe that existing Knowledge Distillation (KD) methods are unsuitable towards pre-training since they normally distill the logits that are going to be discarded when transferred to downstream tasks. To resolve this problem, we propose a feature-based KD method with non-parametric feature dimension aligning. Notably, our method performs comparably with supervised pre-training counterparts in 3 downstream tasks and 9 downstream datasets requiring 10x less data and 5x less pre-training time. Code is available at https://github.com/CVMI-Lab/KDEP.

Learning discriminative deep feature embeddings by using million-scale in-the-wild datasets and margin-based softmax loss is the current state-of-the-art approach for face recognition. However, the memory and computing cost of the Fully Connected (FC) layer linearly scales up to the number of identities in the training set. Besides, the large-scale training data inevitably suffers from inter-class conflict and long-tailed distribution. In this paper, we propose a sparsely updating variant of the FC layer, named Partial FC (PFC). In each iteration, positive class centers and a random subset of negative class centers are selected to compute the margin-based softmax loss. All class centers are still maintained throughout the whole training process, but only a subset is selected and updated in each iteration. Therefore, the computing requirement, the probability of inter-class conflict, and the frequency of passive update on tail class centers, are dramatically reduced. Extensive experiments across different training data and backbones (e.g. CNN and ViT) confirm the effectiveness, robustness and efficiency of the proposed PFC.