CVPR 2024 Papers

The Segment Anything Model (SAM) has garnered significant attention for its versatile segmentation abilities and intuitive prompt-based interface. However its application in medical imaging presents challenges requiring either substantial training costs and extensive medical datasets for full model fine-tuning or high-quality prompts for optimal performance. This paper introduces H-SAM: a prompt-free adaptation of SAM tailored for efficient fine-tuning of medical images via a two-stage hierarchical decoding procedure. In the initial stage H-SAM employs SAM's original decoder to generate a prior probabilistic mask guiding a more intricate decoding process in the second stage. Specifically we propose two key designs: 1) A class-balanced mask-guided self-attention mechanism addressing the unbalanced label distribution enhancing image embedding; 2) A learnable mask cross-attention mechanism spatially modulating the interplay among different image regions based on the prior mask. Moreover the inclusion of a hierarchical pixel decoder in H-SAM enhances its proficiency in capturing fine-grained and localized details. This approach enables SAM to effectively integrate learned medical priors facilitating enhanced adaptation for medical image segmentation with limited samples. Our H-SAM demonstrates a 4.78% improvement in average Dice compared to existing prompt-free SAM variants for multi-organ segmentation using only 10% of 2D slices. Notably without using any unlabeled data H-SAM even outperforms state-of-the-art semi-supervised models relying on extensive unlabeled training data across various medical datasets. Our code is available at https://github.com/Cccccczh404/H-SAM.

Domain adaptation is a critical task in machine learning that aims to improve model performance on a target domain by leveraging knowledge from a related source domain. In this work we introduce Universal Semi-Supervised Domain Adaptation (UniSSDA) a practical yet challenging setting where the target domain is partially labeled and the source and target label space may not strictly match. UniSSDA is at the intersection of Universal Domain Adaptation (UniDA) and Semi-Supervised Domain Adaptation (SSDA): the UniDA setting does not allow for fine-grained categorization of target private classes not represented in the source domain while SSDA focuses on the restricted closed-set setting where source and target label spaces match exactly. Existing UniDA and SSDA methods are susceptible to common-class bias in UniSSDA settings where models overfit to data distributions of classes common to both domains at the expense of private classes. We propose a new prior-guided pseudo-label refinement strategy to reduce the reinforcement of common-class bias due to pseudo-labeling a common label propagation strategy in domain adaptation. We demonstrate the effectiveness of the proposed strategy on benchmark datasets Office-Home DomainNet and VisDA. The proposed strategy attains the best performance across UniSSDA adaptation settings and establishes a new baseline for UniSSDA.

Existing fine-tuning methods for computer vision tasks primarily focus on re-weighting the knowledge learned from the source domain during pre-training. They aim to retain beneficial knowledge for the target domain while suppressing unfavorable knowledge. During the pre-training and fine-tuning stages there is a notable disparity in the data scale. Consequently it is theoretically necessary to employ a model with reduced complexity to mitigate the potential structural risk. However our empirical investigation in this paper reveals that models fine-tuned using existing methods still manifest a high level of model complexity inherited from the pre-training stage leading to a suboptimal stability and generalization ability. This phenomenon indicates an issue that has been overlooked in fine-tuning: Structural Risk Minimization. To address this issue caused by data scale disparity during the fine-tuning stage we propose a simple yet effective approach called Tuning Stable Rank Shrinkage (TSRS). TSRS mitigates the structural risk during the fine-tuning stage by constraining the noise sensitivity of the target model based on stable rank theories. Through extensive experiments we demonstrate that incorporating TSRS into fine-tuning methods leads to improved generalization ability on various tasks regardless of whether the neural networks are based on convolution or transformer architectures. Additionally empirical analysis reveals that TSRS enhances the robustness convexity and smoothness of the loss landscapes in fine-tuned models. Code is available at https://github.com/WitGotFlg/TSRS.

As foundation models become more popular there is a growing need to efficiently finetune them for downstream tasks. Although numerous adaptation methods have been proposed they are designed to be efficient only in terms of how many parameters are trained. They however typically still require backpropagating gradients throughout the model meaning that their training-time and -memory cost does not reduce as significantly. We propose an adaptation method which does not backpropagate gradients through the backbone. We achieve this by designing a lightweight network in parallel that operates on features from the frozen pretrained backbone. As a result our method is efficient not only in terms of parameters but also in training-time and memory usage. Our approach achieves state-of-the-art accuracy-parameter trade-offs on the popular VTAB benchmark and we further show how we outperform prior works with respect to training-time and -memory usage too. We further demonstrate the training efficiency and scalability of our method by adapting a vision transformer backbone of 4 billion parameters for the computationally demanding task of video classification without any intricate model parallelism. Here we outperform a prior adaptor-based method which could only scale to a 1 billion parameter backbone or fully-finetuning a smaller backbone with the same GPU and less training time.

In this paper we first assess and harness various Vision Foundation Models (VFMs) in the context of Domain Generalized Semantic Segmentation (DGSS). Driven by the motivation that Leveraging Stronger pre-trained models and Fewer trainable parameters for Superior generalizability we introduce a robust fine-tuning approach namely "Rein" to parameter-efficiently harness VFMs for DGSS. Built upon a set of trainable tokens each linked to distinct instances Rein precisely refines and forwards the feature maps from each layer to the next layer within the backbone. This process produces diverse refinements for different categories within a single image. With fewer trainable parameters Rein efficiently fine-tunes VFMs for DGSS tasks surprisingly surpassing full parameter fine-tuning. Extensive experiments across various settings demonstrate that Rein significantly outperforms state-of-the-art methods. Remarkably with just an extra 1% of trainable parameters within the frozen backbone Rein achieves a mIoU of 68.1% on the Cityscapes without accessing any real urban-scene datasets. Code is available at https://github.com/w1oves/Rein.git.

Source-Free Domain Adaptation (SFDA) aims to adapt a source model for a target domain with only access to unlabeled target training data and the source model pretrained on a supervised source domain. Relying on pseudo labeling and/or auxiliary supervision conventional methods are inevitably error-prone. To mitigate this limitation in this work we for the first time explore the potentials of off-the-shelf vision-language (ViL) multimodal models (e.g. CLIP) with rich whilst heterogeneous knowledge. We find that directly applying the ViL model to the target domain in a zero-shot fashion is unsatisfactory as it is not specialized for this particular task but largely generic. To make it task specific we propose a novel Distilling multImodal Foundation mOdel (DIFO) approach. Specifically DIFO alternates between two steps during adaptation: (i) Customizing the ViL model by maximizing the mutual information with the target model in a prompt learning manner (ii) Distilling the knowledge of this customized ViL model to the target model. For more fine-grained and reliable distillation we further introduce two effective regularization terms namely most-likely category encouragement and predictive consistency. Extensive experiments show that DIFO significantly outperforms the state-of-the-art alternatives. Code is here.

The pursuit of accurate 3D hand pose estimation stands as a keystone for understanding human activity in the realm of egocentric vision. The majority of existing estimation methods still rely on single-view images as input leading to potential limitations e.g. limited field-of-view and ambiguity in depth. To address these problems adding another camera to better capture the shape of hands is a practical direction. However existing multi-view hand pose estimation methods suffer from two main drawbacks: 1) Requiring multi-view annotations for training which are expensive. 2) During testing the model becomes inapplicable if camera parameters/layout are not the same as those used in training. In this paper we propose a novel Single-to-Dual-view adaptation (S2DHand) solution that adapts a pre-trained single-view estimator to dual views. Compared with existing multi-view training methods 1) our adaptation process is unsupervised eliminating the need for multi-view annotation. 2) Moreover our method can handle arbitrary dual-view pairs with unknown camera parameters making the model applicable to diverse camera settings. Specifically S2DHand is built on certain stereo constraints including pair-wise cross-view consensus and invariance of transformation between both views. These two stereo constraints are used in a complementary manner to generate pseudo-labels allowing reliable adaptation. Evaluation results reveal that S2DHand achieves significant improvements on arbitrary camera pairs under both in-dataset and cross-dataset settings and outperforms existing adaptation methods with leading performance. Project page: https://github.com/ut-vision/S2DHand.

Fine-tuning is the process of extending the training of pre-trained models on specific target tasks thereby significantly enhancing their performance across various applications. However fine-tuning often demands large memory consumption posing a challenge for low-memory devices that some previous memory-efficient fine-tuning methods attempted to mitigate by pruning activations for gradient computation albeit at the cost of significant computational overhead from the pruning processes during training. To address these challenges we introduce PreBackRazor a novel activation pruning scheme offering both computational and memory efficiency through a sparsified backpropagation strategy which judiciously avoids unnecessary activation pruning and storage and gradient computation. Before activation pruning our approach samples a probability of selecting a portion of parameters to freeze utilizing a bandit method for updates to prioritize impactful gradients on convergence. During the feed-forward pass each model layer adjusts adaptively based on parameter activation status obviating the need for sparsification and storage of redundant activations for subsequent backpropagation. Benchmarking on fine-tuning foundation models our approach maintains baseline accuracy across diverse tasks yielding over 20% speedup and around 10% memory reduction. Moreover integrating with an advanced CUDA kernel achieves up to 60% speedup without extra memory costs or accuracy loss significantly enhancing the efficiency of fine-tuning foundation models on memory-constrained devices.

In this paper we delve into the nuanced challenge of tailoring the Segment Anything Models (SAMs) for integration with event data with the overarching objective of attaining robust and universal object segmentation within the event-centric domain. One pivotal issue at the heart of this endeavor is the precise alignment and calibration of embeddings derived from event-centric data such that they harmoniously coincide with those originating from RGB imagery. Capitalizing on the vast repositories of datasets with paired events and RGB images our proposition is to harness and extrapolate the profound knowledge encapsulated within the pre-trained SAM framework. As a cornerstone to achieving this we introduce a multi-scale feature distillation methodology. This methodology rigorously optimizes the alignment of token embeddings originating from event data with their RGB image counterparts thereby preserving and enhancing the robustness of the overall architecture. Considering the distinct significance that token embeddings from intermediate layers hold for higher-level embeddings our strategy is centered on accurately calibrating the pivotal token embeddings. This targeted calibration is aimed at effectively managing the discrepancies in high-level embeddings originating from both the event and image domains. Extensive experiments on different datasets demonstrate the effectiveness of the proposed distillation method. Code in https://github.com/happychenpipi/EventSAM.

In the expansive domain of computer vision a myriad of pre-trained models are at our disposal. However most of these models are designed for natural RGB images and prove inadequate for spectral remote sensing (RS) images. Spectral RS images have two main traits: (1) multiple bands capturing diverse feature information (2) spatial alignment and consistent spectral sequencing within the spatial-spectral dimension. In this paper we introduce Spatial-SpectralMAE (S2MAE) a specialized pre-trained architecture for spectral RS imagery. S2MAE employs a 3D transformer for masked autoencoder modeling integrating learnable spectral-spatial embeddings with a 90% masking ratio. The model efficiently captures local spectral consistency and spatial invariance using compact cube tokens demonstrating versatility to diverse input characteristics. This adaptability facilitates progressive pretraining on extensive spectral datasets. The effectiveness of S2MAE is validated through continuous pretraining on two sizable datasets totaling over a million training images. The pre-trained model is subsequently applied to three distinct downstream tasks with in-depth ablation studies conducted to emphasize its efficacy.

Humans effortlessly interpret images by parsing them into part-whole hierarchies; deep learning excels in learning multi-level feature spaces but they often lack explicit coding of part-whole relations a prominent property of medical imaging. To overcome this limitation we introduce Adam-v2 a new self-supervised learning framework extending Adam [68] by explicitly incorporating part-whole hierarchies into its learning objectives through three key branches: (1) Localizability acquiring discriminative representations to distinguish different anatomical patterns; (2) Composability learning each anatomical structure in a parts-to-whole manner; and (3) Decomposability comprehending each anatomical structure in a whole-to-parts manner. Experimental results across 10 tasks compared to 11 baselines in zero-shot few-shot transfer and full fine-tuning settings showcase Adam-v2's superior performance over large-scale medical models and existing SSL methods across diverse downstream tasks. The higher generality and robustness of Adam-v2's representations originate from its explicit construction of hierarchies for distinct anatomical structures from unlabeled medical images. Adam-v2 preserves a semantic balance of anatomical diversity and harmony in its embedding yielding representations that are both generic and semantically meaningful yet overlooked in existing SSL methods. All code and pretrained models are available at GitHub.com/JLiangLab/Eden.

Developing generalizable manipulation skills is a core challenge in embodied AI. This includes generalization across diverse task configurations encompassing variations in object shape density friction coefficient and external disturbances such as forces applied to the robot. Rapid Motor Adaptation (RMA) offers a promising solution to this challenge. It posits that essential hidden variables influencing an agent's task performance such as object mass and shape can be effectively inferred from the agent's action and proprioceptive history. Drawing inspiration from RMA in locomotion and in-hand rotation we use depth perception to develop agents tailored for rapid motor adaptation in a variety of manipulation tasks. We evaluated our agents on four challenging tasks from the Maniskill2 benchmark namely pick-and-place operations with hundreds of objects from the YCB and EGAD datasets peg insertion with precise position and orientation and operating a variety of faucets and handles with customized environment variations. Empirical results demonstrate that our agents surpass state-of-the-art methods like automatic domain randomization and vision-based policies obtaining better generalization performance and sample efficiency.

Multi-modality large language models (MLLMs) as represented by GPT-4V have introduced a paradigm shift for visual perception and understanding tasks that a variety of abilities can be achieved within one foundation model. While current MLLMs demonstrate primary low-level visual abilities from the identification of low-level visual attributes (e.g. clarity brightness) to the evaluation on image quality there's still an imperative to further improve the accuracy of MLLMs to substantially alleviate human burdens. To address this we collect the first dataset consisting of human natural language feedback on low-level vision. Each feedback offers a comprehensive description of an image's low-level visual attributes culminating in an overall quality assessment. The constructed Q-Pathway dataset includes 58K detailed human feedbacks on 18973 multi-sourced images with diverse low-level appearance. To ensure MLLMs can adeptly handle diverse queries we further propose a GPT-participated transformation to convert these feedbacks into a rich set of 200K instruction-response pairs termed Q-Instruct. Experimental results indicate that the Q-Instruct consistently elevates various low-level visual capabilities across multiple base models. We anticipate that our datasets can pave the way for a future that foundation models can assist humans on low-level visual tasks.

The rapid increase in cases of non-alcoholic fatty liver disease (NAFLD) in recent years has raised significant public concern. Accurately identifying tissue alteration regions is crucial for the diagnosis of NAFLD but this task presents challenges in pathology image analysis particularly with small-scale datasets. Recently the paradigm shift from full fine-tuning to prompting in adapting vision foundation models has offered a new perspective for small-scale data analysis. However existing prompting methods based on task-agnostic prompts are mainly developed for generic image recognition which fall short in providing instructive cues for complex pathology images. In this paper we propose Q uantitative A ttribute-based P rompting (QAP) a novel prompting method specifically for liver pathology image analysis. QAP is based on two quantitative attributes namely K-function-based spatial attributes and histogram-based morphological attributes which are aimed for quantitative assessment of tissue states. Moreover a conditional prompt generator is designed to turn these instance-specific attributes into visual prompts. Extensive experiments on three diverse tasks demonstrate that our task-specific prompting method achieves better diagnostic performance as well as better interpretability. Code is available at \href https://github.com/7LFB/QAP https://github.com/7LFB/QAP .

Recent advances in large-scale pretraining have yielded visual foundation models with strong capabilities. Not only can recent models generalize to arbitrary images for their training task their intermediate representations are useful for other visual tasks such as detection and segmentation. Given that such models can classify delineate and localize objects in 2D we ask whether they also represent their 3D structure? In this work we analyze the 3D awareness of visual foundation models. We posit that 3D awareness implies that representations (1) encode the 3D structure of the scene and (2) consistently represent the surface across views. We conduct a series of experiments using task-specific probes and zero-shot inference procedures on frozen features. Our experiments reveal several limitations of the current models. Our code and analysis can be found at https://github.com/mbanani/probe3d.

Visual-language foundation models like CLIP learn generalized representations that enable zero-shot open-set classification. Few-shot adaptation methods based on prompt tuning have been shown to further improve performance on downstream datasets. However these methods do not fare well in the taxonomic open set (TOS) setting where the classifier is asked to make prediction from label set across different levels of semantic granularity. Frequently they infer incorrect labels at coarser taxonomic class levels even when the inference at the leaf level (original class labels) is correct. To address this problem we propose a prompt tuning technique that calibrates the hierarchical consistency of model predictions. A set of metrics of hierarchical consistency the Hierarchical Consistent Accuracy (HCA) and the Mean Treecut Accuracy (MTA) are first proposed to evaluate TOS model performance. A new Prompt Tuning for Hierarchical Consistency (ProTeCt) technique is then proposed to calibrate classification across label set granularities. Results show that ProTeCt can be combined with existing prompt tuning methods to significantly improve TOS classification without degrading the leaf level classification performance.

Parameter-efficient fine-tuning (PEFT) is an effective methodology to unleash the potential of large foundation models in novel scenarios with limited training data. In the computer vision community PEFT has shown effectiveness in image classification but little research has studied its ability for image segmentation. Fine-tuning segmentation models usually require a heavier adjustment of parameters to align the proper projection directions in the parameter space for new scenarios. This raises a challenge to existing PEFT algorithms as they often inject a limited number of individual parameters into each block which prevents substantial adjustment of the projection direction of the parameter space due to the limitation of Hidden Markov Chain along blocks. In this paper we equip PEFT with a cross-block orchestration mechanism to enable the adaptation of the Segment Anything Model (SAM) to various downstream scenarios. We introduce a novel inter-block communication module which integrates a learnable relation matrix to facilitate communication among different coefficient sets of each PEFT block's parameter space. Moreover we propose an intra-block enhancement module which introduces a linear projection head whose weights are generated from a hyper-complex layer further enhancing the impact of the adjustment of projection directions on the entire parameter space. Extensive experiments on diverse benchmarks demonstrate that our proposed approach consistently improves the segmentation performance significantly on novel scenarios with only around 1K additional parameters.

Customization techniques for text-to-image models have paved the way for a wide range of previously unattainable applications enabling the generation of specific concepts across diverse contexts and styles. While existing methods facilitate high-fidelity customization for individual concepts or a limited pre-defined set of them they fall short of achieving scalability where a single model can seamlessly render countless concepts. In this paper we address a new problem called Modular Customization with the goal of efficiently merging customized models that were fine-tuned independently for individual concepts. This allows the merged model to jointly synthesize concepts in one image without compromising fidelity or incurring any additional computational costs. To address this problem we introduce Orthogonal Adaptation a method designed to encourage the customized models which do not have access to each other during fine-tuning to have orthogonal residual weights. This ensures that during inference time the customized models can be summed with minimal interference. Our proposed method is both simple and versatile applicable to nearly all optimizable weights in the model architecture. Through an extensive set of quantitative and qualitative evaluations our method consistently outperforms relevant baselines in terms of efficiency and identity preservation demonstrating a significant leap toward scalable customization of diffusion models.