CVPR 2025 Papers

The goal of visual grounding is to establish connections between target objects and textual descriptions. Large Language Models (LLMs) have demonstrated strong comprehension abilities across a variety of visual tasks. To establish precise associations between the text and the corresponding visual region, we propose a Task-aware Cross-modal feature Refinement Transformer with LLMs for visual grounding, and our model is referred to as TCRT. To enable the LLM trained solely on text to understand images, we introduce an LLM adaptation module that extracts text-related visual features to bridge the domain discrepancy between the textual and visual modalities. We feed the text and visual features into the LLM to obtain task-aware priors. To enable the priors to guide feature fusion process, we further incorporate a cross-modal feature fusion module, which allows task-aware embeddings to refine visual features and facilitate information interaction between the Referring Expression Comprehension (REC) and Referring Expression Segmentation (RES) tasks. We have performed extensive experiments to verify the effectiveness of the main components and demonstrate the superior performance of the proposed TCRT over state-of-the-art end-to-end visual grounding methods on RefCOCO, RefCOCOg, RefCOCO+ and ReferItGame.

The ability to learn new concepts while preserve the learned knowledge is desirable for learning systems in Class-Incremental Learning (CIL). Recently, feature expansion of the model become a prevalent solution for CIL, where the old features are fixed during the training of the new task while new features are expanded for the new tasks. However, such task-specific features learned from the new task may collide with the old features, leading to misclassification between tasks. Therefore, the expanded model is often encouraged to capture diverse features from the new task, aiming to avoid such collision. However, the existing solution is largely restricted to the samples from the current task, because of the poor accessibility to previous samples. To promote the learning and transferring of diverse features across tasks, we propose a framework called Task-Agnostic Guided Feature Expansion (TagFex). Firstly, it captures task-agnostic features continually with a separate model, providing extra task-agnostic features for subsequent tasks. Secondly, to obtain useful features from the task-agnostic model for the current task, it aggregates the task-agnostic features with the task-specific feature using a merge attention. Then the aggregated feature is transferred back into the task-specific feature for inference, helping the task-specific model capture diverse features. Extensive experiments show the effectiveness and superiority of TagFex on various CIL settings.

The rapid advancement of face forgery techniques has introduced a growing variety of forgeries.Incremental Face Forgery Detection (IFFD), involvinggradually adding new forgery data to fine-tune the previously trained model, has been introduced as a promising strategy to deal with evolving forgery methods.However, a naively trained IFFD model is prone to catastrophic forgetting when new forgeries are integrated, as treating all forgeries as a single "Fake" class in the Real/Fake classification can cause different forgery types overriding one another, thereby resulting in the forgetting of unique characteristics from earlier tasks and limiting the model's effectiveness in learning forgery specificity and generality.In this paper, we propose to stack the latent feature distributions of previous and new tasks brick by brick, i.e., achieving aligned feature isolation. In this manner, we aim to preserve learned forgery information and accumulate new knowledge by minimizing distribution overriding, thereby mitigating catastrophic forgetting.To achieve this, we first introduce Sparse Uniform Replay (SUR) to obtain the representative subsets that could be treated as the uniformly sparse versions of the previous global distributions.We then propose a Latent-space Incremental Detector (LID) that leverages SUR data to isolate and align distributions. For evaluation, we construct a more advanced and comprehensive benchmark tailored for IFFD. The leading experimental results validate the superiority of our method.

Recently, improving the residual structure and designing efficient convolutions have become important branches of lightweight visual reconstruction model design. We have observed that the feature addition mode (FAM) in existing residual structure tends to lead to slow feature learning or stagnation in feature evolution, a phenomenon we define as network inertia. In addition, although blueprint separable convolutions (BSConv) have proved the dominance of intra-kernel correlation, BSConv forces the blueprint to perform scale transformation on all channels, which may lead to incorrect intra-kernel correlation and introduce useless or disruptive features on some channels and hinder the effective propagation of features. Therefore, in this paper, we rethink the FAM and BSConv for super-light visual reconstruction framework design. First, we design a novel linking mode, called feature diversity evolution link (FDEL), which aims to alleviate the phenomenon of network inertia by reducing the retention of previous low-level features, thereby promoting the evolution of feature diversity. Second, we propose blueprint controllable convolutions (B2Conv). The B2Conv can adaptively pick accurate intra-kernel correlation in the depth-axis, effectively preventing the introduction of useless or disruptive features. Based on FDEL and B2Conv, we develop a super-light super-resolution (SR) framework SLVR for visual reconstruction. Both FDEL and B2Conv can serve as efficient plugins. Extensive experimental results demonstrate the effectiveness of our proposed B2Conv, FDEL, and SLVR. Our code will be available at https://github.com/chongningni/SLVR.

Recent advancements in deep learning have greatly enhanced 3D object recognition, but most models are limited to closed-set scenarios, unable to handle unknown samples in real-world applications. Open-set recognition (OSR) addresses this limitation by enabling models to both classify known classes and identify novel classes.However, current OSR methods rely on global features to differentiate known and unknown classes, treating the entire object uniformly and overlooking the varying semantic importance of its different parts.To address this gap, we propose Salience-Aware Structured Separation (SASep), which includes (i) a tunable semantic decomposition (TSD) module to semantically decompose objects into important and unimportant parts, (ii) a geometric synthesis strategy (GSS) to generate pseudo-unknown objects by combining these unimportant parts, and (iii) a synth-aided margin separation (SMS) module to enhance feature-level separation by expanding the feature distributions between classes.Together, these components improve both geometric and feature representations, enhancing the model's ability to effectively distinguish known and unknown classes.Experimental results show that SASep achieves superior performance in 3D OSR, outperforming existing state-of-the-art methods. The codes are available at https://github.com/JinfengX/SASep.

This paper tackles the emerging challenge of multi-view virtual try-on, utilizing both front- and back-view clothing images as inputs. Extending frontal try-on methods to a multi-view context is not straightforward. Simply concatenating the two input views or encoding their features for a generative model, such as a diffusion model, often fails to produce satisfactory results. The main challenge lies in effectively extracting and fusing meaningful clothing features from these input views. Existing explicit warping based methods, which establish direct correspondence between input and target views, tend to introduce artifacts, particularly when there is a significant disparity between the input and target views. Conversely, implicit encoding based methods often lose spatial information about clothing, resulting in outputs that lack detail. To overcome these challenges, we propose Robust-MVTON, an end-to end method for robust and high-quality multi-view try-ons. Our approach introduces a novel cross-pose feature alignment technique to guide the fusion of clothing features and incorporates a newly designed loss function for training. With the fused multi-scale clothing features, we employ a coarse-to-fine diffusion model to generate realistic and detailed results. Extensive experiments conducted on the Deepfashion and MPV datasets affirm the superiority of our method, achieving state-of-the-art performance.

Category-agnostic pose estimation aims to locate keypoints on query images according to a few annotated support images for arbitrary novel classes. Existing methods generally extract support features via heatmap pooling, and obtain interacted features from support and query via cross-attention. Hence, these works neglect to mine fine-grained and structure-aware (FGSA) features from both support and query images, which are crucial for pixel-level keypoint localization. To this end, we propose a novel yet concise framework, which recurrently mines FGSA features from both support and query images. Specifically, we design a FGSA mining module based on deformable attention mechanism. On the one hand, we mine fine-grained features by applying deformable attention head over multi-scale feature maps. On the other hand, we mine structure-aware features by offsetting the reference points of keypoints to their linked keypoints. By means of above module, we recurrently mine FGSA features from support and query images, and thus obtain better support features and query estimations. In addition, we propose to use mixup keypoints to pad various classes to a unified keypoint number, which could provide richer supervision than the zero padding used in existing works. We conduct extensive experiments and in-depth studies on large-scale MP-100 dataset, and outperform SOTA method dramatically (+3.2% PCK\@0.05).

As a core step in structure-from-motion and SLAM, robust feature detection and description under challenging scenarios such as significant viewpoint changes remain unresolved despite their ubiquity. While recent works have identified the importance of local features in modeling geometric transformations, these methods fail to learn the visual cues present in long-range relationships. We present Robust Deformable Detector (RDD), a novel and robust keypoint detector/descriptor leveraging the deformable transformer, which captures global context and geometric invariance through deformable self-attention mechanisms. Specifically, we observed that deformable attention focuses on key locations, effectively reducing the search space complexity and modeling the geometric invariance. Furthermore, we collected an Air-to-Ground dataset for training in addition to the standard MegaDepth dataset. Our proposed method outperforms all state-of-the-art keypoint detection/description methods in sparse matching tasks and is also capable of semi-dense matching. To ensure comprehensive evaluation, we introduce two challenging benchmarks: one emphasizing large viewpoint and scale variations, and the other being an Air-to-Ground benchmark -- an evaluation setting that has recently gaining popularity for 3D reconstruction across different altitudes.

Thermal object detection is a critical task in various fields, such as surveillance and autonomous driving. Current state-of-the-art (SOTA) models always leverage a prior Thermal-To-Visible (T2V) translation model to obtain visible spectrum information, followed by a cross-modality aggregation module to fuse information from both modalities. However, this fusion approach does not fully exploit the complementary visible spectrum information beneficial for thermal detection. To address this issue, we propose a novel cross-modal fusion method called Pseudo Visible Feature Fine-Grained Fusion (PFGF). Specifically, a graph is constructed with nodes generated from multi-level thermal features and pseudo-visual latent features produced by the T2V model. Each level of features corresponds to a subgraph. An Inter-Mamba block is proposed to perform cross-modality fusion between nodes at the lowest level; while a Cascade Knowledge Integration (CKI) strategy is used to fuse low-level fused information to high-level subgraphs in a cascade manner. After several iterations of graph node updating, each subgraph outputs an aggregated feature to the detection head respectively. Unlike previous cross-modal fusion methods, our approach explicitly models high-level relationships between cross-modal data, effectively fusing different granularity information. Experimental results demonstrate that our method achieves SOTA detection performance. Code is available at https://github.com/liting1018/PFGF.

Real-time and high-performance 3D object detection plays a critical role in autonomous driving and robotics. Recent pillar-based 3D object detectors have gained significant attention due to their compact representation and low computational overhead, making them suitable for onboard deployment and quantization. However, existing pillar-based detectors still suffer from information loss along height dimension and large numerical distribution difference during pillar feature encoding (PFE), which severely limits their performance and quantization potential. To address above issue, we first unveil the importance of different input information during PFE and identify the height dimension as a key factor in enhancing 3D detection performance. Motivated by this observation, we propose a height-aware pillar feature encoder, called PillarHist. Specifically, PillarHist statistics the discrete distribution of points at different heights within one pillar. This simple yet effective design greatly preserves the information along the height dimension while significantly reducing the computation overhead of the PFE. Meanwhile, PillarHist also constrains the arithmetic distribution of PFE input to a stable range, making it quantization-friendly. Notably, PillarHist operates exclusively within the PFE stage to enhance performance, enabling seamless integration into existing pillar-based methods without introducing complex operations. Extensive experiments show the effectiveness of PillarHist in terms of both efficiency and performance.

We tackle the problem of automatic calibration of radially distorted cameras in challenging conditions. Accurately determining distortion parameters typically requires either 1) solving the full Structure from Motion (SfM) problem involving camera poses, 3D points, and the distortion parameters, which is only possible if many images with sufficient overlap are provided, or 2) relying heavily on learning-based methods that are comparatively less accurate. In this work, we demonstrate that distortion calibration can be decoupled from 3D reconstruction, maintaining the accuracy of SfM-based methods while avoiding many of the associated complexities. This is achieved by working in Projective Space, where the geometry is unique up to a homography, which encapsulates all camera parameters except for distortion. Our proposed method, Projective Radial Distortion Averaging, averages multiple distortion estimates in a fully projective framework without creating 3d points and full bundle adjustment. By relying on pairwise projective relations, our methods support any feature-matching approaches without constructing point tracks across multiple images.

Adaptive categorization of visual scenes is essential for AI agents to handle changing tasks. Unlike fixed common categories for plants or animals, ad-hoc categories, such as things to sell at a garage sale, are created dynamically to achieve specific tasks. We study open ad-hoc categorization, where the goal is to infer novel concepts and categorize images based on a given context, a small set of labeled exemplars, and some unlabeled data. We have two key insights: 1) recognizing ad-hoc categories relies on the same perceptual processes as common categories; 2) novel concepts can be discovered semantically by expanding contextual cues or visually by clustering similar patterns. We propose OAK, a simple model that introduces a single learnable context token into CLIP, trained with CLIP's objective of aligning visual and textual features and GCD's objective of clustering similar images. On Stanford and Clevr-4 datasets, OAK consistently achieves the state-of-art in accuracy and concept discovery across multiple categorizations, including 87.4% novel accuracy on Stanford Mood, surpassing CLIP and GCD by over 50%. Moreover, OAK generates interpretable saliency maps, focusing on hands for Action, faces for Mood, and backgrounds for Location, promoting transparency and trust while enabling accurate and flexible categorization.

Collaborative perception in autonomous driving significantly enhances the perception capabilities of individual agents. Immutable heterogeneity in collaborative perception, where agents have different and fixed perception networks, presents a major challenge due to the semantic gap in their exchanged intermediate features without modifying the perception networks. Most existing methods bridge the semantic gap through interpreters. However, they either require training a new interpreter for each new agent type, limiting extensibility, or rely on a two-stage interpretation via an intermediate standardized semantic space, causing cumulative semantic loss. To achieve both extensibility in immutable heterogeneous scenarios and low-loss feature interpretation, we propose PolyInter, a polymorphic feature interpreter. It contains an extension point through which emerging new agents can seamlessly integrate by overriding only their specific prompts, which are learnable parameters intended to guide the interpretation, while reusing PolyInter's remaining parameters. By leveraging polymorphism, our design ensures that single interpreter is sufficient to accommodate diverse agents and interpret their features into the ego agent's semantic space. Experiments conducted on OPV2V dataset demonstrate that PolyInter improves collaborative perception precision by up to 11.1% compared to SOTA interpreters, while comparable results can be achieved by training only 1.4% of PolyInter's parameters when adapting to new agents.

Multimodal Large Language Models (MLLMs) have made significant advancements in recent years, with visual features playing an increasingly critical role in enhancing model performance. However, the integration of multi-layer visual features in MLLMs remains underexplored, particularly with regard to optimal layer selection and fusion strategies. Existing methods often rely on arbitrary design choices, leading to suboptimal outcomes. In this paper, we systematically investigate two core aspects of multi-layer visual feature fusion: (1) selecting the most effective visual layers and (2) identifying the best fusion approach with the language model. Our experiments reveal that while combining visual features from multiple stages improves generalization, incorporating additional features from the same stage typically leads to diminished performance. Furthermore, we find that direct fusion of multi-layer visual features at the input stage consistently yields superior and more stable performance across various configurations.

Existing state-of-the-art feature matchers capture long-range dependencies with Transformers but are hindered by high spatial complexity,leading to demanding training and high-latency inference.Striking a better balance between performance and efficiency remains a critical challenge in feature matching.Inspired by the linear complexity \mathcal O (N) of Mamba, we propose an ultra-lightweight Mamba-based matcher, named JamMa, which converges on a single GPU and achieves an impressive performance-efficiency balance in inference.To unlock the potential of Mamba for feature matching,we propose Joint Mamba with a scan-merge strategy named JEGO, which enables:(1) Joint scan of two images to achieve high-frequency mutual interaction, (2) Efficient scan with skip steps to reduce sequence length, (3) Global receptive field, and (4) Omnidirectional feature representation.With the above properties, the JEGO strategy significantly outperforms the scan-merge strategies proposed in VMamba and EVMamba in the feature matching task.Compared to attention-based sparse and semi-dense matchers, JamMa demonstrates a notably superior balance between performance and efficiency,delivering better performance with less than 50% of the parameters and FLOPs.

Semi-supervised semantic segmentation with consistencyregularization capitalizes on unlabeled images to enhancethe accuracy of pixel-level segmentation. Current consistencylearning methods primarily rely on the consistency loss be-tween pseudo-labels and unlabeled images, neglecting the in-formation within the feature representations of the backboneencoder. Preserving maximum information in feature embed-dings requires achieving the alignment and uniformity objec-tives, as widely studied. To address this, we present SWSEG,a semi-supervised semantic segmentation algorithm that opti-mizes alignment and uniformity using the Sliced-WassersteinDistance (SWD), and rigorously and empirically proves thisconnection. We further resolve the computational issues as-sociated with conventional Monte Carlo-based SWD by im-plementing a Gaussian-approximated variant, which not onlymaintains the alignment and uniformity objectives but alsoimproves training efficiency. We evaluate SWSEG on thePASCAL VOC 2012, Cityscapes, and ADE20K datasets, out-shining supervised baselines in mIoU by up to 11.8%, 8.9%,and 8.2%, respectively, given an equivalent number of labeledsamples. Further, SWSEG surpasses state-of-the-art methodsin multiple settings across these three datasets. Our extensiveablation studies confirm the optimization of the uniformityand alignment objectives of the feature representations.

Person re-identification (ReID) aims to extract accurate identity representation features. However, during feature extraction, individual samples are inevitably affected by noise (background, occlusions, and model limitations). Considering that features from the same identity follow a normal distribution around identity centers after training, we propose a Training-Free Feature Centralization ReID framework (Pose2ID) by aggregating the same identity features to reduce individual noise and enhance the stability of identity representation, which preserves the feature's original distribution for following strategies such as re-ranking. Specifically, to obtain samples of the same identity, we introduce two components: Identity-Guided Pedestrian Generation: by leveraging identity features to guide the generation process, we obtain high-quality images with diverse poses, ensuring identity consistency even in complex scenarios such as infrared, and occlusion. Neighbor Feature Centralization: it explores each sample's potential positive samples from its neighborhood. Experiments demonstrate that our generative model exhibits strong generalization capabilities and maintains high identity consistency. With the Feature Centralization framework, we achieve impressive performance even with an ImageNet pre-trained model without ReID training, reaching mAP/Rank-1 of 52.81/78.92 on Market1501. Moreover, our method sets new state-of-the-art results across standard, cross-modality, and occluded ReID tasks, showcasing strong adaptability.

Normal integration reconstructs 3D surfaces from normal maps obtained, e.g. by photometric stereo. These normal maps capture surface details down to the pixel level but require large computational resources for integration at high resolutions. In this work, we replace the dense pixel grid with a sparse anisotropic triangle mesh prior to normal integration. We adapt the triangle mesh to the local geometry in the case of complex surface structures and remove oversampling from flat featureless regions. For high-resolution images, the resulting compression reduces normal integration runtimes from hours to minutes while maintaining high surface accuracy. Our main contribution is the derivation of the well-known quadric error measure from mesh decimation for screen space applications and its combination with optimal Delaunay triangulation.