CVPR 2024 Papers

Knowledge distillation is an effective method for training small and efficient deep learning models. However the efficacy of a single method can degenerate when transferring to other tasks modalities or even other architectures. To address this limitation we propose a novel constrained feature distillation method. This method is derived from a small set of core principles which results in two emerging components: an orthogonal projection and a task-specific normalisation. Equipped with both of these components our transformer models can outperform all previous methods on ImageNet and reach up to a 4.4% relative improvement over the previous state-of-the-art methods. To further demonstrate the generality of our method we apply it to object detection and image generation whereby we obtain consistent and substantial performance improvements over state-of-the-art. Code and models are publicly available.

Teeth localization segmentation and labeling in 2D images have great potential in modern dentistry to enhance dental diagnostics treatment planning and population-based studies on oral health. However general instance segmentation frameworks are incompetent due to 1) the subtle differences between some teeth' shapes (e.g. maxillary first premolar and second premolar) 2) the teeth's position and shape variation across subjects and 3) the presence of abnormalities in the dentition (e.g. caries and edentulism). To address these problems we propose a ViT-based framework named TeethSEG which consists of stacked Multi-Scale Aggregation (MSA) blocks and an Anthropic Prior Knowledge (APK) layer. Specifically to compose the two modules we design 1) a unique permutation-based upscaler to ensure high efficiency while establishing clear segmentation boundaries with 2) multi-head self/cross-gating layers to emphasize particular semantics meanwhile maintaining the divergence between token embeddings. Besides we collect 3) the first open-sourced intraoral image dataset IO150K which comprises over 150k intraoral photos and all photos are annotated by orthodontists using a human-machine hybrid algorithm. Experiments on IO150K demonstrate that our TeethSEG outperforms the state-of-the-art segmentation models on dental image segmentation.

Despite extensive research on training generative adversarial networks (GANs) with limited training data learning to generate images from long-tailed training distributions remains fairly unexplored. In the presence of imbalanced multi-class training data GANs tend to favor classes with more samples leading to the generation of low quality and less diverse samples in tail classes. In this study we aim to improve the training of class-conditional GANs with long-tailed data. We propose a straightforward yet effective method for knowledge sharing allowing tail classes to borrow from the rich information from classes with more abundant training data. More concretely we propose modifications to existing class-conditional GAN architectures to ensure that the lower-resolution layers of the generator are trained entirely unconditionally while reserving class-conditional generation for the higher-resolution layers. Experiments on several long-tail benchmarks and GAN architectures demonstrate a significant improvement over existing methods in both the diversity and fidelity of the generated images. The code is available at https://github.com/khorrams/utlo.

Data-free knowledge distillation is able to utilize the knowledge learned by a large teacher network to augment the training of a smaller student network without accessing the original training data avoiding privacy security and proprietary risks in real applications. In this line of research existing methods typically follow an inversion-and-distillation paradigm in which a generative adversarial network on-the-fly trained with the guidance of the pre-trained teacher network is used to synthesize a large-scale sample set for knowledge distillation. In this paper we reexamine this common data-free knowledge distillation paradigm showing that there is considerable room to improve the overall training efficiency through a lens of "small-scale inverted data for knowledge distillation". In light of three empirical observations indicating the importance of how to balance class distributions in terms of synthetic sample diversity and difficulty during both data inversion and distillation processes we propose Small Scale Data-free Knowledge Distillation (SSD-KD). In formulation SSD-KD introduces a modulating function to balance synthetic samples and a priority sampling function to select proper samples facilitated by a dynamic replay buffer and a reinforcement learning strategy. As a result SSD-KD can perform distillation training conditioned on an extremely small scale of synthetic samples (e.g. 10x less than the original training data scale) making the overall training efficiency one or two orders of magnitude faster than many mainstream methods while retaining superior or competitive model performance as demonstrated on popular image classification and semantic segmentation benchmarks. The code is available at https://github.com/OSVAI/SSD-KD.

Despite the recent success of sign language research the widely adopted CNN-based backbones are mainly migrated from other computer vision tasks in which the contours and texture of objects are crucial for identifying objects. They usually treat sign frames as grids and may fail to capture effective cross-region features. In fact sign language tasks need to focus on the correlation of different regions in one frame and the interaction of different regions among adjacent frames for identifying a sign sequence. In this paper we propose to represent a sign sequence as graphs and introduce a simple yet effective graph-based sign language processing architecture named SignGraph to extract cross-region features at the graph level. SignGraph consists of two basic modules: Local Sign Graph (LSG) module for learning the correlation of intra-frame cross-region features in one frame and Temporal Sign Graph (TSG) module for tracking the interaction of inter-frame cross-region features among adjacent frames. With LSG and TSG we build our model in a multiscale manner to ensure that the representation of nodes can capture cross-region features at different granularities. Extensive experiments on current public sign language datasets demonstrate the superiority of our SignGraph model. Our model achieves very competitive performances with the SOTA model while not using any extra cues. Code and models are available at: https://github.com/gswycf/SignGraph.

Reasoning from visual dynamics scenes has many real world applications. However existing video reasoning benchmarks are still inadequate since they were mainly designed for factual or situated reasoning and rarely involve broader knowledge in the real world. Our work aims to delve deeper into reasoning evaluations specifically within dynamic open-world and structured context knowledge. We propose a new benchmark (SOK-Bench) consisting of 44K questions and 10K situations with instance-level annotations depicted in the videos. The reasoning process is required to understand and apply situated knowledge and general knowledge for problem-solving. To create such a dataset we propose an automatic and scalable generation method to generate question-answer pairs knowledge graphs and rationales by instructing the combinations of LLMs and MLLMs. Concretely we first extract observable situated entities relations and processes from videos for situated knowledge and then extend to open-world knowledge beyond the visible content. The task generation is facilitated through multiple dialogues as iterations and subsequently corrected and refined by our designed self-promptings and demonstrations. With a corpus of both explicit situated facts and implicit commonsense we generate associated question-answer pairs and reasoning processes finally followed by manual reviews for quality assurance. We evaluated recent mainstream large vision language models on the benchmark and found several insightful conclusions. For more information please refer to our benchmark at www.bobbywu.com/SOKBench.

Scene graphs have been recently introduced into 3D spatial understanding as a comprehensive representation of the scene. The alignment between 3D scene graphs is the first step of many downstream tasks such as scene graph aided point cloud registration mosaicking overlap checking and robot navigation. In this work we treat 3D scene graph alignment as a partial graph-matching problem and propose to solve it with a graph neural network. We reuse the geometric features learned by a point cloud registration method and associate the clustered point-level geometric features with the node-level semantic feature via our designed feature fusion module. Partial matching is enabled by using a learnable method to select the top-k similar node pairs. Subsequent downstream tasks such as point cloud registration are achieved by running a pre-trained registration network within the matched regions. We further propose a point-matching rescoring method that uses the node-wise alignment of the 3D scene graph to reweight the matching candidates from a pre-trained point cloud registration method. It reduces the false point correspondences estimated especially in low-overlapping cases. Experiments show that our method improves the alignment accuracy by 10 20% in low-overlap and random transformation scenarios and outperforms the existing work in multiple downstream tasks. Our code and models are available here (https://github.com/dfki-av/sg-pgm.git).

The inherent noisy and sparse characteristics of radar data pose challenges in finding effective representations for 3D object detection. In this paper we propose RadarDistill a novel knowledge distillation (KD) method which can improve the representation of radar data by leveraging LiDAR data. RadarDistill successfully transfers desirable characteristics of LiDAR features into radar features using three key components: Cross-Modality Alignment (CMA) Activation-based Feature Distillation (AFD) and Proposal-based Feature Distillation (PFD). CMA enhances the density of radar features by employing multiple layers of dilation operations effectively addressing the challenge of inefficient knowledge transfer from LiDAR to radar. AFD selectively transfers knowledge based on regions of the LiDAR features with a specific focus on areas where activation intensity exceeds a predefined threshold. PFD similarly guides the radar network to selectively mimic features from the LiDAR network within the object proposals. Our comparative analyses conducted on the nuScenes datasets demonstrate that RadarDistill achieves state-of-the-art (SOTA) performance for radar-only object detection task recording 20.5% in mAP and 43.7% in NDS. Also RadarDistill significantly improves the performance of the camera-radar fusion model.

The goal of Universal Cross-Domain Retrieval (UCDR) is to achieve robust performance in generalized test scenarios wherein data may belong to strictly unknown domains and categories during training. Recently pre-trained models with prompt tuning have shown strong generalization capabilities and attained noteworthy achievements in various downstream tasks such as few-shot learning and video-text retrieval. However applying them directly to UCDR may not be sufficient to handle both domain shift (i.e. adapting to unfamiliar domains) and semantic shift (i.e. transferring to unknown categories). To this end we propose Prompting-to-Simulate (ProS) the first method to apply prompt tuning for UCDR. ProS employs a two-step process to simulate Content-aware Dynamic Prompts (CaDP) which can impact models to produce generalized features for UCDR. Concretely in Prompt Units Learning stage we introduce two Prompt Units to individually capture domain and semantic knowledge in a mask-and-align way. Then in Context-aware Simulator Learning stage we train a Content-aware Prompt Simulator under a simulated test scenario to produce the corresponding CaDP. Extensive experiments conducted on three benchmark datasets show that our method achieves new state-of-the-art performance without bringing excessive parameters. Code is available at https://github.com/fangkaipeng/ProS

With the rapidly increasing demand for oriented object detection (OOD) recent research involving weakly-supervised detectors for learning rotated box (RBox) from the horizontal box (HBox) has attracted more and more attention. In this paper we explore a more challenging yet label-efficient setting namely single point-supervised OOD and present our approach called Point2RBox. Specifically we propose to leverage two principles: 1) Synthetic pattern knowledge combination: By sampling around each labeled point on the image we spread the object feature to synthetic visual patterns with known boxes to provide the knowledge for box regression. 2) Transform self-supervision: With a transformed input image (e.g. scaled/rotated) the output RBoxes are trained to follow the same transformation so that the network can perceive the relative size/rotation between objects. The detector is further enhanced by a few devised techniques to cope with peripheral issues e.g. the anchor/layer assignment as the size of the object is not available in our point supervision setting. To our best knowledge Point2RBox is the first end-to-end solution for point-supervised OOD. In particular our method uses a lightweight paradigm yet it achieves a competitive performance among point-supervised alternatives 41.05%/27.62%/80.01% on DOTA/DIOR/HRSC datasets.

Foundation models encompass an extensive knowledge base and offer remarkable transferability. However this knowledge becomes outdated or insufficient over time. The challenge lies in continuously updating foundation models to accommodate novel information while retaining their original capabilities. Leveraging the fact that foundation models have initial knowledge on various tasks and domains we propose a novel approach that instead of updating all parameters equally localizes the updates to a sparse set of parameters relevant to the task being learned. We strike a balance between efficiency and new task performance while maintaining the transferability and generalizability of foundation models. We extensively evaluate our method on foundational vision-language models with a diverse spectrum of continual learning tasks. Our method achieves improvements on the accuracy of the newly learned tasks up to 7% while preserving the pretraining knowledge with a negligible decrease of 0.9% on a representative control set accuracy.

Current approaches for 3D scene graph prediction rely on labeled datasets to train models for a fixed set of known object classes and relationship categories. We present Open3DSG an alternative approach to learn 3D scene graph prediction in an open world without requiring labeled scene graph data. We co-embed the features from a 3D scene graph prediction backbone with the feature space of powerful open world 2D vision language foundation models. This enables us to predict 3D scene graphs from 3D point clouds in a zero-shot manner by querying object classes from an open vocabulary and predicting the inter-object relationships from a grounded LLM with scene graph features and queried object classes as context. Open3DSG is the first 3D point cloud method to predict not only explicit open-vocabulary object classes but also open-set relationships that are not limited to a predefined label set making it possible to express rare as well as specific objects and relationships in the predicted 3D scene graph. Our experiments show that Open3DSG is effective at predicting arbitrary object classes as well as their complex inter-object relationships describing spatial supportive semantic and comparative relationships.

We estimate the radiance field of large-scale dynamic areas from multiple vehicle captures under varying environmental conditions. Previous works in this domain are either restricted to static environments do not scale to more than a single short video or struggle to separately represent dynamic object instances. To this end we present a novel decomposable radiance field approach for dynamic urban environments. We propose a multi-level neural scene graph representation that scales to thousands of images from dozens of sequences with hundreds of fast-moving objects. To enable efficient training and rendering of our representation we develop a fast composite ray sampling and rendering scheme. To test our approach in urban driving scenarios we introduce a new novel view synthesis benchmark. We show that our approach outperforms prior art by a significant margin on both established and our proposed benchmark while being faster in training and rendering.

The scarcity of annotated data has sparked significant interest in unsupervised pre-training methods that leverage medical reports as auxiliary signals for medical visual representation learning. However existing research overlooks the multi-granularity nature of medical visual representation and lacks suitable contrastive learning techniques to improve the models' generalizability across different granularities leading to the underutilization of image-text information. To address this we propose MLIP a novel framework leveraging domain-specific medical knowledge as guiding signals to integrate language information into the visual domain through image-text contrastive learning. Our model includes global contrastive learning with our designed divergence encoder local token-knowledge-patch alignment contrastive learning and knowledge-guided category-level contrastive learning with expert knowledge. Experimental evaluations reveal the efficacy of our model in enhancing transfer performance for tasks such as image classification object detection and semantic segmentation. Notably MLIP surpasses state-of-the-art methods even with limited annotated data highlighting the potential of multimodal pre-training in advancing medical representation learning.

The popularity of large-scale pre-training has promoted the development of medical foundation models. However some studies have shown that although foundation models exhibit strong general feature extraction capabilities their performance on specific tasks is still inferior to task-specific methods. In this paper we explore a new perspective called "Knowledge Decomposition" to improve the performance on specific medical tasks which deconstruct the foundation model into multiple lightweight expert models each dedicated to a particular task with the goal of improving specialization while concurrently mitigating resource expenditure. To accomplish the above objective we design a novel framework named Low-Rank Knowledge Decomposition (LoRKD) which explicitly separates graidents by incorporating low-rank expert modules and the efficient knowledge separation convolution. Extensive experimental results demonstrate that the decomposed models perform well in terms of performance and transferability even surpassing the original foundation models. Source code is available at: https://github.com/MediaBrain-SJTU/LoRKD

Knowledge distillation involves transferring soft labels from a teacher to a student using a shared temperature-based softmax function. However the assumption of a shared temperature between teacher and student implies a mandatory exact match between their logits in terms of logit range and variance. This side-effect limits the performance of student considering the capacity discrepancy between them and the finding that the innate logit relations of teacher are sufficient for student to learn. To address this issue we propose setting the temperature as the weighted standard deviation of logit and performing a plug-and-play Z-score pre-process of logit standardization before applying softmax and Kullback-Leibler divergence. Our pre-process enables student to focus on essential logit relations from teacher rather than requiring a magnitude match and can improve the performance of existing logit-based distillation methods. We also show a typical case where the conventional setting of sharing temperature between teacher and student cannot reliably yield the authentic distillation evaluation; nonetheless this challenge is successfully alleviated by our Z-score. We extensively evaluate our method for various student and teacher models on CIFAR-100 and ImageNet showing its significant superiority. The vanilla knowledge distillation powered by our pre-process can achieve favorable performance against state-of-the-art methods and other distillation variants can obtain considerable gain with the assistance of our pre-process. The codes pre-trained models and logs are released on Github.

We present the Locally Adaptive Morphable Model (LAMM) a highly flexible Auto-Encoder (AE) framework for learning to generate and manipulate 3D meshes. We train our architecture following a simple self-supervised training scheme in which input displacements over a set of sparse control vertices are used to overwrite the encoded geometry in order to transform one training sample into another. During inference our model produces a dense output that adheres locally to the specified sparse geometry while maintaining the overall appearance of the encoded object. This approach results in state-of-the-art performance in both disentangling manipulated geometry and 3D mesh reconstruction. To the best of our knowledge LAMM is the first end-to-end framework that enables direct local control of 3D vertex geometry in a single forward pass. A very efficient computational graph allows our network to train with only a fraction of the memory required by previous methods and run faster during inference generating 12k vertex meshes at >60fps on a single CPU thread. We further leverage local geometry control as a primitive for higher level editing operations and present a set of derivative capabilities such as swapping and sampling object parts. Code and pretrained models can be found at https://github.com/michaeltrs/LAMM.

The goal of the multi-sound source localization task is to localize sound sources from the mixture individually. While recent multi-sound source localization methods have shown improved performance they face challenges due to their reliance on prior information about the number of objects to be separated. In this paper to overcome this limitation we present a novel multi-sound source localization method that can perform localization without prior knowledge of the number of sound sources. To achieve this goal we propose an iterative object identification (IOI) module which can recognize sound-making objects in an iterative manner. After finding the regions of sound-making objects we devise object similarity-aware clustering (OSC) loss to guide the IOI module to effectively combine regions of the same object but also distinguish between different objects and backgrounds. It enables our method to perform accurate localization of sound-making objects without any prior knowledge. Extensive experimental results on the MUSIC and VGGSound benchmarks show the significant performance improvements of the proposed method over the existing methods for both single and multi-source. Our code is available at: https://github.com/VisualAIKHU/NoPrior_MultiSSL

Geometric knowledge has been shown to be beneficial for the stereo matching task. However prior attempts to integrate geometric insights into stereo matching algorithms have largely focused on geometric knowledge from single images while crucial cross-view factors such as occlusion and matching uniqueness have been overlooked. To address this gap we propose a novel Intra-view and Cross-view Geometric knowledge learning Network (ICGNet) specifically crafted to assimilate both intra-view and cross-view geometric knowledge. ICGNet harnesses the power of interest points to serve as a channel for intra-view geometric understanding. Simultaneously it employs the correspondences among these points to capture cross-view geometric relationships. This dual incorporation empowers the proposed ICGNet to leverage both intra-view and cross-view geometric knowledge in its learning process substantially improving its ability to estimate disparities. Our extensive experiments demonstrate the superiority of the ICGNet over contemporary leading models. The code will be available at https://github.com/DFSDDDDD1199/ICGNet.