CVPR 2025 | Papers-with-Code | 【合集四】检测与分割(目前已更15篇,持续更新中)

在这里插入图片描述
CVPR 2025 decisions are now available on OpenReview!22.1% = 2878 / 13008
会议官网:https://cvpr.thecvf.com/Conferences/2025

目前计划整理六个合集,部分合集未发布
【合集一】AIGC
【合集二】Mamba、MLLM
【合集三】底层视觉
【合集四】检测与分割
【合集五】三维视觉
【合集六】视频理解

欢迎转载,转载注明出处哦
——————————————————————————————————————————————————————————————

OOD 检测

1.《DPU: Dynamic Prototype Updating for Multimodal Out-of-Distribution Detection》
Paper: https://arxiv.org/pdf/2411.08227
Code: https://github.com/lili0415/DPU-OOD-Detection
在这里插入图片描述

2.《OODD: Test-time Out-of-Distribution Detection with Dynamic Dictionary》
Paper: https://arxiv.org/pdf/2503.10468
Code: https://github.com/zxk1212/OODD
在这里插入图片描述

目标检测

1.《MonoDGP: Monocular 3D Object Detection with Decoupled-Query and Geometry-Error Priors》
Paper:https://arxiv.org/pdf/2410.19590
Code:https://github.com/PuFanqi23/MonoDGP
在这里插入图片描述

2.《V2X-R: Cooperative LiDAR-4D Radar Fusion with Denoising Diffusion for 3D Object Detection》
Paper: https://arxiv.org/pdf/2411.08402
Code: https://github.com/ylwhxht/V2X-R
在这里插入图片描述

3.《FASTer: Focal Token Acquiring-and-Scaling Transformerfor Long-term 3D Object Detection》
Paper: https://arxiv.org/pdf/2503.01899
Code: https://github.com/MSunDYY/FASTer
在这里插入图片描述

4.《Object Detection using Event Camera: A MoE Heat Conduction based Detector and A New Benchmark Dataset》
Paper: https://arxiv.org/pdf/2412.06647
Code: https://github.com/Event-AHU/OpenEvDET
在这里插入图片描述

5.《Multi-Sensor Object Anomaly Detection:Unifying Appearance, Geometry, and Internal Properties》
Paper: https://arxiv.org/pdf/2412.14592
Code: https://github.com/ZZZBBBZZZ/MulSen-AD
在这里插入图片描述

6.《Point2RBox-v2: Rethinking Point-supervised Oriented Object Detection with Spatial Layout Among Instances》
Paper: https://arxiv.org/pdf/2502.04268
Code: https://github.com/VisionXLab/point2rbox-v2
在这里插入图片描述

7.《LLMDet: Learning Strong Open-Vocabulary Object Detectors under the Supervision of Large Language Models》
Paper: https://arxiv.org/pdf/2501.18954
Code: https://github.com/iSEE-Laboratory/LLMDet
在这里插入图片描述

8.《MonoTAKD: Teaching Assistant Knowledge Distillation for Monocular 3DObject Detection》
Paper: https://arxiv.org/pdf/2404.04910
Code: https://github.com/hoiliu-0801/MonoTAKD
在这里插入图片描述

语义分割

1.《SegMAN: Omni-scale Context Modeling with State Space Modelsand Local Attention for Semantic Segmentation》
Paper: https://arxiv.org/pdf/2412.11890
Code: https://github.com/yunxiangfu2001/SegMAN
在这里插入图片描述

2.《An End-to-End Robust Point Cloud Semantic SegmentationNetwork with Single-Step Conditional Diffusion Models》
Paper: https://arxiv.org/pdf/2411.16308
Code: https://github.com/QWTforGithub/CDSegNet
在这里插入图片描述

3.《Exact: Exploring Space-Time Perceptive Clues for Weakly SupervisedSatellite Image Time Series Semantic Segmentation》
Paper: https://arxiv.org/pdf/2412.03968
Code: https://github.com/MiSsU-HH/Exact
在这里插入图片描述

4.《BFANet: Revisiting 3D Semantic Segmentation with Boundary Feature Analysis》
Paper:https://arxiv.org/pdf/2503.12539
Code:https://github.com/weiguangzhao/BFANet
在这里插入图片描述

5.《Golden Cudgel Network for Real-Time Semantic Segmentation》
Paper: https://arxiv.org/pdf/2503.03325
Code: https://github.com/gyyang23/GCNet
在这里插入图片描述

### CVPR 2025 Conference Object Detection Papers Summary The CVPR (Conference on Computer Vision and Pattern Recognition) is a premier annual event that showcases advancements in computer vision research. For the year 2025, numerous contributions have been made specifically within the domain of object detection. #### Advances in Semi-Supervised Learning for 3D Object Detection from RGB-D Data One significant contribution involves leveraging semi-supervised learning techniques to enhance 3D object detection using RGB-D data[^1]. This approach aims at improving model robustness by utilizing both labeled and unlabeled datasets effectively. The method introduces novel strategies for transferring knowledge across different environments while maintaining high accuracy levels even when training samples are scarce or partially annotated. #### Evaluation Metrics for Saliency Datasets Applied to Object Detection Another critical aspect highlighted during this conference pertains to evaluating models' performance through specialized metrics designed particularly for saliency-based tasks such as object detection. Two primary measures—`Object/Area Ratio` and `Contrast Score`—have emerged as essential indicators of an algorithm's effectiveness in identifying objects accurately under varying conditions[^2]. ```python def calculate_object_area_ratio(obj_area, total_image_area): """ Calculate the ratio between object area and image total area. Args: obj_area (float): Area covered by detected objects. total_image_area (float): Total area of input images. Returns: float: Ratio value indicating how much space objects occupy relative to entire scene. """ return obj_area / total_image_area def compute_contrast_score(object_intensity, background_intensity): """ Compute contrast score based on intensity differences between foreground & background regions Args: object_intensity (list[float]): List containing pixel intensities inside bounding boxes. background_intensity (list[float]): Corresponding list outside those areas but still close enough Returns: float: A measure representing visibility against surroundings which helps assess detectability quality. """ avg_obj_int = sum(object_intensity)/len(object_intensity) avg_bg_int = sum(background_intensity)/len(background_intensity) return abs(avg_obj_int - avg_bg_int) ``` These developments reflect ongoing efforts towards refining methodologies used not only in detecting targets more precisely but also ensuring these systems can generalize well beyond controlled settings into real-world applications where they will be most beneficial.
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值