CVPR 2022 Papers

We present CLIP-NeRF, a multi-modal 3D object manipulation method for neural radiance fields (NeRF). By leveraging the joint language-image embedding space of the recent Contrastive Language-Image Pre-Training (CLIP) model, we propose a unified framework that allows manipulating NeRF in a user-friendly way, using either a short text prompt or an exemplar image. Specifically, to combine the novel view synthesis capability of NeRF and the controllable manipulation ability of latent representations from generative models, we introduce a disentangled conditional NeRF architecture that allows individual control over both shape and appearance. This is achieved by performing the shape conditioning via applying a learned deformation field to the positional encoding and deferring color conditioning to the volumetric rendering stage. To bridge this disentangled latent representation to the CLIP embedding, we design two code mappers that take a CLIP embedding as input and update the latent codes to reflect the targeted editing. The mappers are trained with a CLIP-based matching loss to ensure the manipulation accuracy. Furthermore, we propose an inverse optimization method that accurately projects an input image to the latent codes for manipulation to enable editing on real images. We evaluate our approach by extensive experiments on a variety of text prompts and exemplar images and also provide an intuitive editing interface for real-time user interaction.

Generating shapes using natural language can enable new ways of imagining and creating the things around us. While significant recent progress has been made in text-to-image generation, text-to-shape generation remains a challenging problem due to the unavailability of paired text and shape data at a large scale. We present a simple yet effective method for zero-shot text-to-shape generation that circumvents such data scarcity. Our proposed method, named CLIP-Forge, is based on a two-stage training process, which only depends on an unlabelled shape dataset and a pre-trained image-text network such as CLIP. Our method has the benefits of avoiding expensive inference time optimization, as well as the ability to generate multiple shapes for a given text. We not only demonstrate promising zero-shot generalization of the CLIP-Forge model qualitatively and quantitatively, but also provide extensive comparative evaluations to better understand its behavior.

It has been widely known that CAM (Class Activation Map) usually only activates discriminative object regions and falsely includes lots of object-related backgrounds. As only a fixed set of image-level object labels are available to the WSSS (weakly supervised semantic segmentation) model, it could be very difficult to suppress those diverse background regions consisting of open set objects. In this paper, we propose a novel Cross Language Image Matching (CLIMS) framework, based on the recently introduced Contrastive Language-Image Pre-training (CLIP) model, for WSSS. The core idea of our framework is to introduce natural language supervision to activate more complete object regions and suppress closely-related open background regions. In particular, we design object, background region and text label matching losses to guide the model to excite more reasonable object regions for CAM of each category. In addition, we design a co-occurring background suppression loss to prevent the model from activating closely-related background regions, with a predefined set of class-related background text descriptions. These designs enable the proposed CLIMS to generate a more complete and compact activation map for the target objects. Extensive experiments on PASCAL VOC2012 dataset show that our CLIMS significantly outperforms the previous state-of-the-art methods.

Channel pruning has been broadly recognized as an effective technique to reduce the computation and memory cost of deep convolutional neural networks. However, conventional pruning methods have limitations in that: they are restricted to pruning process only, and they require a fully pre-trained large model. Such limitations may lead to sub-optimal model quality as well as excessive memory and training cost. In this paper, we propose a novel Channel Exploration methodology, dubbed as CHEX, to rectify these problems. As opposed to pruning-only strategy, we propose to repeatedly prune and regrow the channels throughout the training process, which reduces the risk of pruning important channels prematurely. More exactly: From intra-layer's aspect, we tackle the channel pruning problem via a well-known column subset selection (CSS) formulation. From inter-layer's aspect, our regrowing stages open a path for dynamically re-allocating the number of channels across all the layers under a global channel sparsity constraint. In addition, all the exploration process is done in a single training from scratch without the need of a pre-trained large model. Experimental results demonstrate that CHEX can effectively reduce the FLOPs of diverse CNN architectures on a variety of computer vision tasks, including image classification, object detection, instance segmentation, and 3D vision. For example, our compressed ResNet-50 model on ImageNet dataset achieves 76% top-1 accuracy with only 25% FLOPs of the original ResNet-50 model, outperforming previous state-of-the-art channel pruning methods. The checkpoints and code are available at here.

The objective of human parsing is to partition a human in an image into constituent parts. This task involves labeling each pixel of the human image according to the classes. Since the human body comprises hierarchically structured parts, each body part of an image can have its sole position distribution characteristic. Probably, a human head is less likely to be under the feet, and arms are more likely to be near the torso. Inspired by this observation, we make instance class distributions by accumulating the original human parsing label in the horizontal and vertical directions, which can be utilized as supervision signals. Using these horizontal and vertical class distribution labels, the network is guided to exploit the intrinsic position distribution of each class. We combine two guided features to form a spatial guidance map, which is then superimposed onto the baseline network by multiplication and concatenation to distinguish the human parts precisely. We conducted extensive experiments to demonstrate the effectiveness and superiority of our method on three well-known benchmarks: LIP, ATR, and CIHP databases.

Federated learning (FL) is a distributed learning paradigm that enables multiple clients to collaboratively learn a shared global model. Despite the recent progress, it remains challenging to deal with heterogeneous data clients, as the discrepant data distributions usually prevent the global model from delivering good generalization ability on each participating client. In this paper, we propose CD^2-pFed, a novel Cyclic Distillation-guided Channel Decoupling framework, to personalize the global model in FL, under various settings of data heterogeneity. Different from previous works which establish layer-wise personalization to overcome the non-IID data across different clients, we make the first attempt at channel-wise assignment for model personalization, referred to as channel decoupling. To further facilitate the collaboration between private and shared weights, we propose a novel cyclic distillation scheme to impose a consistent regularization between the local and global model representations during the federation. Guided by the cyclical distillation, our channel decoupling framework can deliver more accurate and generalized results for different kinds of heterogeneity, such as feature skew, label distribution skew, and concept shift. Comprehensive experiments on four benchmarks, including natural image and medical image analysis tasks, demonstrate the consistent effectiveness of our method on both local and external validations.

In autonomous driving, LiDAR point-clouds and RGB images are two major data modalities with complementary cues for 3D object detection. However, it is quite difficult to sufficiently use them, due to large inter-modal discrepancies. To address this issue, we propose a novel framework, namely Contrastively Augmented Transformer for multi-modal 3D object Detection (CAT-Det). Specifically, CAT-Det adopts a two-stream structure consisting of a Pointformer (PT) branch, an Imageformer (IT) branch along with a Cross-Modal Transformer (CMT) module. PT, IT and CMT jointly encode intra-modal and inter-modal long-range contexts for representing an object, thus fully exploring multi-modal information for detection. Furthermore, we propose an effective One-way Multi-modal Data Augmentation (OMDA) approach via hierarchical contrastive learning at both the point and object levels, significantly improving the accuracy only by augmenting point-clouds, which is free from complex generation of paired samples of the two modalities. Extensive experiments on the KITTI benchmark show that CAT-Det achieves a new state-of-the-art, highlighting its effectiveness.

We introduce CAPRI-Net, a self-supervised neural network for learning compact and interpretable implicit representations of 3D computer-aided design (CAD) models, in the form of adaptive primitive assemblies. Given an input 3D shape, our network reconstructs it by an assembly of quadric surface primitives via constructive solid geometry (CSG) operations. Without any ground-truth shape assemblies, our self-supervised network is trained with a reconstruction loss, leading to faithful 3D reconstructions with sharp edges and plausible CSG trees. While the parametric nature of CAD models does make them more predictable locally, at the shape level, there is much structural and topological variation, which presents a significant generalizability challenge to state-of-the-art neural models for 3D shapes. Our network addresses this challenge by adaptive training with respect to each test shape, with which we fine-tune the network that was pre-trained on a model collection. We evaluate our learning framework on both ShapeNet and ABC, the largest and most diverse CAD dataset to date, in terms of reconstruction quality, sharp edges, compactness, and interpretability, to demonstrate superiority over current alternatives for neural CAD reconstruction.

Dataset condensation aims at reducing the network training effort through condensing a cumbersome training set into a compact synthetic one. State-of-the-art approaches largely rely on learning the synthetic data by matching the gradients between the real and synthetic data batches. Despite the intuitive motivation and promising results, such gradient-based methods, by nature, easily overfit to a biased set of samples that produce dominant gradients, and thus lack a global supervision of data distribution. In this paper, we propose a novel scheme to Condense dataset by Aligning FEatures (CAFE), which explicitly attempts to preserve the real-feature distribution as well as the discriminant power of the resulting synthetic set, lending itself to strong generalization capability to various architectures. At the heart of our approach is an effective strategy to align features from the real and synthetic data across various scales, while accounting for the classification of real samples. Our scheme is further backed up by a novel dynamic bi-level optimization, which adaptively adjusts parameter updates to prevent over-/under-fitting. We validate the proposed CAFE across various datasets, and demonstrate that it generally outperforms the state of the art: on the SVHN dataset, for example, the performance gain is up to 11%. Extensive experiments and analysis verify the effectiveness and necessity of proposed designs.

Few-shot classification is a challenging problem that aims to learn a model that can adapt to unseen classes given a few labeled samples. Recent approaches pre-train a feature extractor, and then fine-tune for episodic meta-learning. Other methods leverage spatial features to learn pixel-level correspondence while jointly training a classifier. However, results using such approaches show marginal improvements. In this paper, inspired by the transformer style self-attention mechanism, we propose a strategy to cross-attend and re-weight discriminative features for few-shot classification. Given a base representation of support and query images after global pooling, we introduce a single shared module that projects features and cross-attends in two aspects: (i) query to support, and (ii) support to query. The module computes attention scores between features to produce an attention pooled representation of features in the same class that is later added to the original representation followed by a projection head. This effectively re-weights features in both aspects (i & ii) to produce features that better facilitate improved metric-based meta-learning. Extensive experiments on public benchmarks show our approach outperforms state-of-the-art methods by 3% 5%.

While class activation map (CAM) generated by image classification network has been widely used for weakly supervised object localization (WSOL) and semantic segmentation (WSSS), such classifiers usually focus on discriminative object regions. In this paper, we propose Contrastive learning for Class-agnostic Activation Map (C^2AM) generation only using unlabeled image data, without the involvement of image-level supervision. The core idea comes from the observation that i) semantic information of foreground objects usually differs from their backgrounds; ii) foreground objects with similar appearance or background with similar color/texture have similar representations in the feature space. We form the positive and negative pairs based on the above relations and force the network to disentangle foreground and background with a class-agnostic activation map using a novel contrastive loss. As the network is guided to discriminate cross-image foreground-background, the class-agnostic activation maps learned by our approach generate more complete object regions. We successfully extracted from C^2AM class-agnostic object bounding boxes for object localization and background cues to refine CAM generated by classification network for semantic segmentation. Extensive experiments on CUB-200-2011, ImageNet-1K, and PASCAL VOC2012 datasets show that both WSOL and WSSS can benefit from the proposed C^2AM.

Long-tail object detection suffers from poor performance on tail categories. We reveal that the real culprit lies in the extremely imbalanced distribution of the classifier's weight norm. For conventional softmax cross-entropy loss, such imbalanced weight norm distribution yields ill conditioned decision boundary for categories which have small weight norms. To get rid of this situation, we choose to maximize the cosine similarity between the learned feature and the weight vector of target category rather than the inner-product of them. The decision boundary between any two categories is the angular bisector of their weight vectors. Whereas, the absolutely equal decision boundary is suboptimal because it reduces the model's sensitivity to various categories. Intuitively, categories with rich data diversity should occupy a larger area in the classification space while categories with limited data diversity should occupy a slightly small space. Hence, we devise a Category-Aware Angular Margin Loss (C2AM Loss) to introduce an adaptive angular margin between any two categories. Specifically, the margin between two categories is proportional to the ratio of their classifiers' weight norms. As a result, the decision boundary is slightly pushed towards the category which has a smaller weight norm. We conduct comprehensive experiments on LVIS dataset. C2AM Loss brings 4.9 5.2 AP improvements on different detectors and backbones compared with baseline.

Recently, many excellent weakly supervised semantic segmentation (WSSS) works are proposed based on class activation mapping (CAM). However, there are few works that consider the characteristics of medical images. In this paper, we find that there are mainly two challenges of medical images in WSSS: i) the boundary of object foreground and background is not clear; ii) the co-occurrence phenomenon is very severe in training stage. We thus propose a Causal CAM (C-CAM) method to overcome the above challenges. Our method is motivated by two cause-effect chains including category-causality chain and anatomy-causality chain. The category-causality chain represents the image content (cause) affects the category (effect). The anatomy-causality chain represents the anatomical structure (cause) affects the organ segmentation (effect). Extensive experiments were conducted on three public medical image data sets. Our C-CAM generates the best pseudo masks with the DSC of 77.26%, 80.34% and 78.15% on ProMRI, ACDC and CHAOS compared with other CAM-like methods. The pseudo masks of C-CAM are further used to improve the segmentation performance for organ segmentation tasks. Our C-CAM achieves DSC of 83.83% on ProMRI and DSC of 87.54% on ACDC, which outperforms state-of-the-art WSSS methods. Our code is available at https://github.com/Tian-lab/C-CAM.

We present a learning-based framework, recurrent transformer network (RTN), to restore heavily degraded old films. Instead of performing frame-wise restoration, our method is based on the hidden knowledge learned from adjacent frames that contain abundant information about the occlusion, which is beneficial to restore challenging artifacts of each frame while ensuring temporal coherency. Moreover, contrasting the representation of the current frame and the hidden knowledge makes it possible to infer the scratch position in an unsupervised manner, and such defect localization generalizes well to real-world degradations. To better resolve mixed degradation and compensate for the flow estimation error during frame alignment, we propose to leverage more expressive transformer blocks for spatial restoration. Experiments on both synthetic dataset and real-world old films demonstrate the significant superiority of the proposed RTN over existing solutions. In addition, the same framework can effectively propagate the color from keyframes to the whole video, ultimately yielding compelling restored films.

In Class Incremental Learning (CIL), a classification model is progressively trained at each incremental step on an evolving dataset of new classes, while at the same time, it is required to preserve knowledge of all the classes observed so far. Prototypical representations can be leveraged to model feature distribution for the past data and inject information of former classes in later incremental steps without resorting to stored exemplars. However, if not updated, those representations become increasingly outdated as the incremental learning progresses with new classes. To address the aforementioned problems, we propose a framework which aims to (i) model the semantic drift by learning the relationship between representations of past and novel classes among incremental steps, and (ii) estimate the feature drift, defined as the evolution of the representations learned by models at each incremental step. Semantic and feature drifts are then jointly exploited to infer up-to-date representations of past classes (evanescent representations), and thereby infuse past knowledge into incremental training. We experimentally evaluate our framework achieving exemplar-free SotA results on multiple benchmarks. In the ablation study, we investigate nontrivial relationships between evanescent representations and models.

Most existing works in vision-and-language navigation (VLN) focus on either discrete or continuous environments, training agents that cannot generalize across the two. Although learning to navigate in continuous spaces is closer to the real-world, training such an agent is significantly more difficult than training an agent in discrete spaces. However, recent advances in discrete VLN are challenging to translate to continuous VLN due to the domain gap. The fundamental difference between the two setups is that discrete navigation assumes prior knowledge of the connectivity graph of the environment, so that the agent can effectively transfer the problem of navigation with low-level controls to jumping from node to node with high-level actions by grounding to an image of a navigable direction. To bridge the discrete-to-continuous gap, we propose a predictor to generate a set of candidate waypoints during navigation, so that agents designed with high-level actions can be transferred to and trained in continuous environments. We refine the connectivity graph of Matterport3D to fit the continuous Habitat-Matterport3D, and train the waypoints predictor with the refined graphs to produce accessible waypoints at each time step. Moreover, we demonstrate that the predicted waypoints can be augmented during training to diversify the views and paths, and therefore enhance agent's generalization ability. Through extensive experiments we show that agents navigating in continuous environments with predicted waypoints perform significantly better than agents using low-level actions, which reduces the absolute discrete-to-continuous gap by 11.76% Success Weighted by Path Length (SPL) for the Cross-Modal Matching Agent and 18.24% SPL for the Recurrent VLN-BERT. Our agents, trained with a simple imitation learning objective, outperform previous methods by a large margin, achieving new state-of-the-art results on the testing environments of the R2R-CE and the RxR-CE datasets.

Weakly supervised object localization aims to find a target object region in a given image with only weak supervision, such as image-level labels. Most existing methods use a class activation map (CAM) to generate a localization map; however, a CAM identifies only the most discriminative parts of a target object rather than the entire object region. In this work, we find the gap between classification and localization in terms of the misalignment of the directions between an input feature and a class-specific weight. We demonstrate that the misalignment suppresses the activation of CAM in areas that are less discriminative but belong to the target object. To bridge the gap, we propose a method to align feature directions with a class-specific weight. The proposed method achieves a state-of-the-art localization performance on the CUB-200-2011 and ImageNet-1K benchmarks.

Bridging global context interactions correctly is important for high-fidelity image completion with large masks. Previous methods attempting this via deep or large receptive field (RF) convolutions cannot escape from the dominance of nearby interactions, which may be inferior. In this paper, we propose to treat image completion as a directionless sequence-to-sequence prediction task, and deploy a transformer to directly capture long-range dependence. Crucially, we employ a restrictive CNN with small and non-overlapping RF for weighted token representation, which allows the transformer to explicitly model the long-range visible context relations with equal importance in all layers, without implicitly confounding neighboring tokens when larger RFs are used. To improve appearance consistency between visible and generated regions, a novel attention-aware layer (AAL) is introduced to better exploit distantly related high-frequency features. Overall, extensive experiments demonstrate superior performance compared to state-of-the-art methods on several datasets. Code is available at https://github.com/lyndonzheng/TFill.