CVPR 2025 论文列表

With its powerful visual-language alignment capability, CLIP performs well in zero-shot and few-shot learning tasks. However, we found in experiments that CLIP's logits suffer from serious inter-class confusion problems in downstream tasks, and the ambiguity between categories seriously affects the accuracy. To address this challenge, we propose a novel method called Logits DeConfusion, which effectively learns and eliminates inter-class confusion in logits by combining our Multi-level Adapter Fusion (MAF) module with our Inter-Class Deconfusion (ICD) module. Our MAF extracts features from different levels and fuses them uniformly to enhance feature representation. Our ICD learnably eliminates inter-class confusion in logits with a residual structure. Experimental results show that our method can significantly improve the classification performance and alleviate the inter-class confusion problem. The code is available at https://github.com/LiShuo1001/LDC.

Problems in differentiable rendering often involve optimizing scene parameters that cause motion in image space. The gradients for such parameters tend to be sparse, leading to poor convergence. While existing methods address this sparsity through proxy gradients such as topological derivatives or lagrangian derivatives, they make simplifying assumptions about rendering. Multi-resolution image pyramids offer an alternative approach but prove unreliable in practice. We introduce a method that uses locally orderless images --- where each pixel maps to a histogram of intensities that preserves local variations in appearance. Using an inverse rendering objective that minimizes histogram distance, our method extends support for sparsely defined image gradients and recovers optimal parameters. We validate our method on various inverse problems using both synthetic and real data.

Localizing events in videos based on semantic queries is a pivotal task in video understanding research and user-oriented applications like video search. Yet, current research predominantly relies on natural language queries (NLQs), overlooking the potential of using multimodal queries (MQs) that incorporate images to flexibly represent semantic queries, particularly when it is difficult to express non-verbal or unfamiliar concepts in words. To bridge this gap, we introduce ICQ, a new benchmark designed for localizing events in videos with MQs, alongside an evaluation dataset ICQ-Highlight. To adapt and reevaluate existing video localization models for this new task, we propose 3 Multimodal Query Adaptation methods and a novel Surrogate Fine-tuning strategy, serving as strong baseline methods. ICQ systematically benchmarks 12 state-of-the-art backbone models, spanning from specialized video localization models to Video Large Language Models. Our extensive experiments highlight the high potential of using MQs in real-world applications. We believe this is a first step toward video event localization with MQs.

Fine-tuning based concept erasing has demonstrated promising results in preventing generation of harmful contents from text-to-image diffusion models by removing target concepts while preserving remaining concepts. To maintain the generation capability of diffusion models after concept erasure, it is necessary to remove only the image region containing the target concept when it locally appears in an image, leaving other regions intact. However, prior arts often compromise fidelity of the other image regions in order to erase the localized target concept appearing in a specific area, thereby reducing the overall performance of image generation. To address these limitations, we first introduce a framework called localized concept erasure, which allows for the deletion of only the specific area containing the target concept in the image while preserving the other regions. As a solution for the localized concept erasure, we propose a training-free approach, dubbed Gated Low-rank adaptation for Concept Erasure (GLoCE), that injects a lightweight module into the diffusion model. GLoCE consists of low-rank matrices and a simple gate, determined only by several generation steps for concepts without training. By directly applying GLoCE to image embeddings and designing the gate to activate only for target concepts, GLoCE can selectively remove only the region of the target concepts, even when target and remaining concepts coexist within an image. Extensive experiments demonstrated GLoCE not only improves the image fidelity to text prompts after erasing the localized target concepts, but also outperforms prior arts in efficacy, specificity, and robustness by large margin and can be extended to mass concept erasure.

Recent methods for zero-shot Human-Object Interaction (HOI) detection typically leverage the generalization ability of large Vision-Language Model (VLM), i.e., CLIP, on unseen categories, showing impressive results on various zero-shot settings. However, existing methods struggle to adapt CLIP representations for human-object pairs, as CLIP tends to overlook fine-grained information necessary for distinguishing interactions. To address this issue, we devise, LAIN, a novel zero-shot HOI detection framework designed to enhance the locality and interaction awareness of CLIP representations. The locality awareness, which involves capturing fine-grained details and the spatial structure of individual objects, is achieved by aggregating the information and spatial priors of adjacent neighborhood patches. The interaction awareness, which involves identifying whether and how a human is interacting with an object, is achieved by capturing the interaction pattern between the human and the object. By infusing locality and interaction awareness into CLIP representation, LAIN captures detailed information about the human-object pairs. Our extensive experiments on existing benchmarks show that LAIN outperforms previous methods in various zero-shot settings, demonstrating the importance of locality and interaction awareness for effective zero-shot HOI detection.

We present LoTUS, a novel Machine Unlearning (MU) method that eliminates the influence of training samples from pre-trained models, avoiding retraining from scratch. LoTUS smooths the prediction probabilities of the model up to an information-theoretic bound, mitigating its over-confidence stemming from data memorization. We evaluate LoTUS on Transformer and ResNet18 models against eight baselines across five public datasets. Beyond established MU benchmarks, we evaluate unlearning on ImageNet1k, a large-scale dataset, where retraining is impractical, simulating real-world conditions. Moreover, we introduce the novel Retrain-Free Jensen-Shannon Divergence (RF-JSD) metric to enable evaluation under real-world conditions. The experimental results show that LoTUS outperforms state-of-the-art methods in terms of both efficiency and effectiveness. Code: https://github.com/cspartalis/LoTUS

While Multimodal Large Language Models (MLLMs) excel at generalizing across modalities and tasks, effectively adapting them to specific downstream tasks while simultaneously retaining both general and specialized knowledge remains challenging. Although Low-Rank Adaptation (LoRA) is widely used to efficiently acquire specialized knowledge in MLLMs, it introduces substantial harmful redundancy during visual instruction tuning, which exacerbates the forgetting of general knowledge and degrades downstream task performance.To address this issue, we propose LoRASculpt to eliminate harmful redundant parameters, thereby harmonizing general and specialized knowledge.Specifically, under theoretical guarantees, we introduce sparse updates into LoRA to discard redundant parameters effectively. Furthermore, we propose a Conflict Mitigation Regularizer to refine the update trajectory of LoRA, mitigating knowledge conflicts with the pretrained weights.Extensive experimental results demonstrate that even at very high degree of sparsity (\le 5%), our method simultaneously enhances generalization and downstream task performance. This confirms that our approach effectively mitigates the catastrophic forgetting issue and further promotes knowledge harmonization in MLLMs.

Recent advances in text-to-image customization have enabled high-fidelity, context-rich generation of personalized images, allowing specific concepts to appear in a variety of scenarios. However, current methods struggle with combining multiple personalized models, often leading to attribute entanglement or requiring separate training to preserve concept distinctiveness. We present LoRACLR, a novel approach for multi-concept image generation that merges multiple LoRA models, each fine-tuned for a distinct concept, into a single, unified model without additional individual fine-tuning. LoRACLR uses a contrastive objective to align and merge the weight spaces of these models, ensuring compatibility while minimizing interference. By enforcing distinct yet cohesive representations for each concept, LoRACLR enables efficient, scalable model composition for high-quality, multi-concept image synthesis. Our results highlight the effectiveness of LoRACLR in accurately merging multiple concepts, advancing the capabilities of personalized image generation.

In continual learning (CL), catastrophic forgetting often arises due to feature drift. This challenge is particularly prominent in the exemplar-free continual learning (EFCL) setting, where samples from previous tasks cannot be retained, making it difficult to preserve prior knowledge. To address this issue, some EFCL methods aim to identify feature spaces that minimize the impact on previous tasks while accommodating new ones. However, they rely on static features or outdated statistics stored from old tasks, which prevents them from capturing the dynamic evolution of the feature space in CL, leading to performance degradation over time. In this paper, we introduce the Drift-Resistant Space (DRS), which effectively handles feature drifts without requiring explicit feature modeling or the storage of previous tasks. A novel parameter-efficient fine-tuning approach called Low-Rank Adaptation Subtraction (LoRA^-) is proposed to develop the DRS. This method subtracts the LoRA weights of old tasks from the initial pre-trained weight before processing new task data to establish the DRS for model training. Therefore, LoRA^- enhances stability, improves efficiency, and simplifies implementation. Furthermore, stabilizing feature drifts allows for better plasticity by learning with a triplet loss. Our method consistently achieves state-of-the-art results, especially for long task sequences, across multiple datasets.

Large Language Models (LLMs) such as ChatGPT demonstrate strong few-shot adaptability without requiring fine-tuning, positioning them ideal for data-limited and real-time applications. However, this adaptability has not yet been replicated in current Visual Foundation Models (VFMs), which require explicit fine-tuning with sufficient tuning data. Besides, the pretraining-finetuning paradigm has led to the surge of numerous task-specific modular components, such as Low-Rank Adaptation (LoRA). For the first time, we explore the potential of reusing diverse pre-tuned LoRAs without accessing their original training data, to achieve tuning-free few-shot adaptation in VFMs. Our framework, LoRA Recycle, distills a meta-LoRA from diverse pre-tuned LoRAs with a meta-learning objective, using synthetic data inversely generated from pre-tuned LoRAs themselves. The VFM, once equipped with the meta-LoRA, is empowered to solve new few-shot tasks in a single forward pass, akin to the in-context learning of LLMs. Additionally, we incorporate a double-efficient mechanism, accelerating the data-generation and meta-training process while maintaining or even improving performance. Extensive experiments across various few-shot classification benchmarks across both in- and cross-domain scenarios demonstrate the superiority of our framework. Code is available at https://github.com/Egg-Hu/LoRA-Recycle.

'Pre-training + fine-tuning' has been widely used in various downstream tasks. Parameter-efficient fine-tuning (PEFT) has demonstrated higher efficiency and promising performance comapred to traditional full-tuning. The widely used adapter-based and prompt-based methods in PEFT can be uniformly represented as adding an MLP structure to the pre-trained model. These methods are prone to over-fitting in downstream tasks, due to the difference in data scale and distribution. To address this issue, we propose a new adapter-based PEFT module, i.e., LoKi, which consists of an encoder, a learnable activation layer, and a decoder. To maintain the simplicity of LoKi, we use single-layer linear networks for the encoder and decoder, and for the learnable activation layer, we use a Kolmogorov-Arnold Network (KAN) with the minimal number of layers (only 2 KAN linear layers). With a bottleneck rate much lower than that of Adapter, LoKi is equipped with fewer parameters (only half of Adapter) and eliminates slow training speed and high memory usage of KAN. We conduct extensive experiments on LoKi under image classification and video action recognition across 9 datasets. LoKi demonstrates highly competitive generalization performance compared to other PEFT methods with fewer tunable parameters, ensuring both effectiveness and efficiency.

Recent video large language models (Video LLMs) often depend on costly human annotations or proprietary APIs (e.g., GPT-4o) to produce training data, which limits their training at scale. In this paper, we explore large-scale training for Video LLM with cheap automatic speech recognition (ASR) transcripts. Specifically, we propose a novel streaming training approach that densely interleaves the ASR words and video frames according to their timestamps. Compared to previous studies in vision-language representation with ASR, our method naturally fits the streaming characteristics of ASR, thus enabling the model to learn temporally-aligned, fine-grained vision-language modeling. To support the training algorithm, we introduce a data pipeline for YouTube videos and their closed captions (CC), resulting in \texttt Live-CC-10M pre-training set and \texttt Live-WhisperX-408K high-quality supervised fine-tuning (SFT) set. Remarkably, even without SFT, the pre-trained model \texttt LiveCC-7B demonstrates significant improvements in general video QA and exhibits a new capability in real-time video commentary. To evaluate this, we carefully design a new benchmark \texttt LiveSports-3K , using LLM-as-a-judge to measure the free-form commentary. Experiments show our final model \texttt LiveCC-7B can surpass LLaVA-Video-72B in commentary quality even working in a real-time mode. Meanwhile, it achieves state-of-the-art results at the 7B scale on popular benchmarks such as VideoMME, demonstrating its broad generalizability. All resources of this paper have been released at \href https://showlab.github.io/livecc showlab.github.io/livecc .

Unsupervised domain adaptation (UDA) aims to learn discriminative features from a labeled source domain by supervised learning and to transfer the knowledge to an unlabeled target domain via distribution alignment. However, in some real-world scenarios, e.g., public safety or access control, it's difficult to obtain sufficient source domain data, which hinders the application of existing UDA methods. To this end, this paper investigates a realistic but rarely studied problem called one-shot unsupervised domain adaptation (OSUDA), where there is only one example per category in the source domain and abundant unlabeled samples in the target domain. Compared with UDA, OSUDA faces dual challenges in both feature learning and domain alignment due to the lack of sufficient source data. To address these challenges, we propose a simple but effective link-based contrastive learning (LCL) method for OSUDA. On the one hand, with the help of in-domain links that indicate whether two samples are from the same cluster, LCL can learn discriminative features with abundant unlabeled target data. On the other hand, by constructing cross-domain links that show whether two clusters are bidirectionally matched, LCL can realize accurate domain alignment with only one source sample per category. Extensive experiments conducted on 4 public domain adaptation benchmarks, including VisDA-2017, Office-31, Office-Home, and DomainNet, demonstrate the effectiveness of the proposed LCL under the OSUDA setting. In addition, we build a realistic OSUDA surveillance video face recognition dataset, where LCL consistently improves the recognition accuracy across various face recognition methods.

Fast 3D clothed human reconstruction from monocular video remains a significant challenge in computer vision, particularly in balancing computational efficiency with reconstruction quality. Current approaches are either focused on static image reconstruction but too computationally intensive, or achieve high quality through per-video optimization that requires minutes to hours of processing, making them unsuitable for real-time applications. To this end, we present TemPoFast3D, a novel method that leverages temporal coherency of human appearance to reduce redundant computation while maintaining reconstruction quality. Our approach is a "plug-and play" solution that uniquely transforms pixel-aligned reconstruction networks to handle continuous video streams by maintaining and refining a canonical appearance representation through efficient coordinate mapping. Extensive experiments demonstrate that TemPoFast3D matches or exceeds state-of-the-art methods across standard metrics while providing high-quality textured reconstruction across diverse pose and appearance, with a maximum speed of 12 FPS.

Text images are unique in their dual nature, encompassing both visual and linguistic information. The visual component encompasses structural and appearance-based features, while the linguistic dimension incorporates contextual and semantic elements. In scenarios with degraded visual quality, linguistic patterns serve as crucial supplements for comprehension, highlighting the necessity of integrating both aspects for robust scene text recognition (STR). Contemporary STR approaches often use language models or semantic reasoning modules to capture linguistic features, typically requiring large-scale annotated datasets. Self-supervised learning, which lacks annotations, presents challenges in disentangling linguistic features related to the global context. Typically, sequence contrastive learning emphasizes the alignment of local features, while masked image modeling (MIM) tends to exploit local structures to reconstruct visual patterns, resulting in limited linguistic knowledge. In this paper, we propose a Linguistics-aware Masked Image Modeling (LMIM) approach, which channels the linguistic information into the decoding process of MIM through a separate branch. Specifically, we design a linguistics alignment module to extract vision-independent features as linguistic guidance using inputs with different visual appearances. As features extend beyond mere visual structures, LMIM must consider the global context to achieve reconstruction. Extensive experiments on various benchmarks quantitatively demonstrate our state-of-the-art performance, and attention visualizations qualitatively show the simultaneous capture of both visual and linguistic information. The code is available at https://github.com/zhangyifei01/LMIM.

Recent years, learned image compression has made tremendous progress to achieve impressive coding efficiency. Its coding gain mainly comes from non-linear neural network-based transform and learnable entropy modeling. However, most studies focus on a strong backbone, and few studies consider a low complexity design. In this paper, we propose LALIC, a linear attention modeling for learned image compression. Specially, we propose to use Bi-RWKV blocks, by utilizing the Spatial Mix and Channel Mix modules to achieve more compact feature extraction, and apply the Conv based Omni-Shift module to adapt to two-dimensional latent representation. Furthermore, we propose a RWKV-based Spatial-Channel ConTeXt model (RWKV-SCCTX), that leverages the Bi-RWKV to modeling the correlation between neighboring features effectively. To our knowledge, our work is the first work to utilize efficient Bi-RWKV models with linear attention for learned image compression. Experimental results demonstrate that our method achieves competitive RD performances by outperforming VTM-9.1 by -15.26%, -15.41%, -17.63% in BD-rate on Kodak, CLIC and Tecnick datasets. The code is available at https://github.com/sjtu-medialab/RwkvCompress .

Image rendering from line drawings is vital in design and image generation technologies reduce costs, yet professional line drawings demand preserving complex details. Text prompts struggle with accuracy, and image translation struggles with consistency and fine-grained control. We present LineArt, a framework that transfers complex appearance onto detailed design drawings, facilitating design and artistic creation. It generates high-fidelity appearance while preserving structural accuracy by simulating hierarchical visual cognition and integrating human artistic experience to guide the diffusion process. LineArt overcomes the limitations of current methods in terms of difficulty in fine-grained control and style degradation in design drawings. It requires no precise 3D modeling, physical property specifications, or network training, making it more convenient for design tasks. LineArt consists of two stages: a multi-frequency lines fusion module to supplement the input design drawing with detailed structural information and a two-part painting process for Base Layer Shaping and Surface Layer Coloring. We also present a new design drawing dataset, ProLines, for evaluation. The experiments show that LineArt performs better in accuracy, realism, and material precision compared to SOTAs.

Text-to-video generation enhances content creation but is highly computationally intensive: The computational cost of Diffusion Transformers (DiTs) scales quadratically in the number of pixels. This makes minute-length video generation extremely expensive, limiting most existing models to generating videos of only 10-20 seconds length. We propose a Linear-complexity text-to-video Generation (LinGen) framework whose cost scales linearly in the number of pixels. For the first time, LinGen enables high-resolution minute-length video generation on a single GPU without compromising quality. It replaces the computationally-dominant and quadratic-complexity block, self-attention, with a linear-complexity block called MATE, which consists of an MA-branch and a TE-branch. The MA-branch targets short-to-long-range correlations, combining a bidirectional Mamba2 block with our token rearrangement method, Rotary Major Scan, and our review tokens developed for long video generation. The TE-branch is a novel TEmporal Swin Attention block that focuses on temporal correlations between adjacent tokens and medium-range tokens. The MATE block addresses the adjacency preservation issue of Mamba and improves the consistency of generated videos significantly. Experimental results show that LinGen outperforms DiT (with a 75.6% win rate) in video quality with up to 15x(11.5x) FLOPs (latency) reduction. Furthermore, both automatic metrics and human evaluation demonstrate our LinGen-4B yields comparable video quality to state-of-the-art models (with a 50.5%, 52.1%, 49.1% win rate with respect to Gen-3, LumaLabs, and Kling, respectively). This paves the way to hour-length movie generation and real-time interactive video generation. Project website: https://lineargen.github.io/.

Scene coordinate regression achieves impressive results in outdoor LiDAR localization but requires days of training. Since training needs to be repeated for each new scene, long training times make these impractical for applications requiring time-sensitive system upgrades, such as autonomous driving, drones, robotics, etc. We identify large coverage areas and vast amounts of data in large-scale outdoor scenes as key challenges that limit fast training. In this paper, we propose LightLoc, the first method capable of efficiently learning localization in a new scene at light speed. Beyond freezing the scene-agnostic feature backbone and training only the scene-specific prediction heads, we introduce two novel techniques to address these challenges. First, we introduce sample classification guidance to assist regression learning, reducing ambiguity from similar samples and improving training efficiency. Second, we propose redundant sample downsampling to remove well-learned frames during training, reducing training time without compromising accuracy. In addition, the fast training and confidence estimation characteristics of sample classification enable its integration into SLAM, effectively eliminating error accumulation. Extensive experiments on large-scale outdoor datasets demonstrate that LightLoc achieves state-of-the-art performance with just 1 hour of training--50x faster than existing methods. Our Code is available at https://github.com/liw95/LightLoc.