敬礼
发表于 2025-3-30 12:01:13
http://reply.papertrans.cn/25/2424/242334/242334_51.png
史前
发表于 2025-3-30 12:29:11
,Embedding-Free Transformer with Inference Spatial Reduction for Efficient Semantic Segmentation, state-of-the-art performance with the efficient computation compared to the existing transformer-based semantic segmentation models in three public benchmarks, including ADE20K, Cityscapes and COCO-Stuff. Furthermore, our ISR method reduces the computational cost by up to 61% with minimal mIoU perf
sterilization
发表于 2025-3-30 19:10:38
,VeCLIP: Improving CLIP Training via Visual-Enriched Captions,ive pipeline, we effortlessly scale our dataset up to 300 million samples named VeCap dataset. Our results show significant advantages in image-text alignment and overall model performance. For example, VeCLIP achieves up to . gain in COCO and Flickr30k retrieval tasks under the 12M setting. For dat
肮脏
发表于 2025-3-30 23:07:33
http://reply.papertrans.cn/25/2424/242334/242334_54.png
腐蚀
发表于 2025-3-31 03:42:08
,Learning Representations from Foundation Models for Domain Generalized Stereo Matching,opose a cosine-constrained concatenation cost (C4) space to construct cost volumes. We integrate FormerStereo with state-of-the-art (SOTA) stereo matching networks and evaluate its effectiveness on multiple benchmark datasets. Experiments show that the FormerStereo framework effectively improves the
Arboreal
发表于 2025-3-31 08:13:04
,Spike-Temporal Latent Representation for Energy-Efficient Event-to-Video Reconstruction,esholding Algorithm. Then, the U-shape SNN decoder reconstructs the video based on the encoded spikes. Experimental results demonstrate that the STLR achieves performance comparable to popular SNNs on IJRR, HQF, and MVSEC datasets while significantly enhancing energy efficiency.
monologue
发表于 2025-3-31 10:44:04
http://reply.papertrans.cn/25/2424/242334/242334_57.png
Torrid
发表于 2025-3-31 17:02:54
,Chat-Edit-3D: Interactive 3D Scene Editing via Text Prompts,rmore, we design a scheme utilizing Hash-Atlas to represent 3D scene views, which transfers the editing of 3D scenes onto 2D atlas images. This design achieves complete decoupling between the 2D editing and 3D reconstruction processes, enabling . to flexibly integrate a wide range of existing 2D or
爵士乐
发表于 2025-3-31 21:12:09
http://reply.papertrans.cn/25/2424/242334/242334_59.png
alcoholism
发表于 2025-4-1 01:15:28
,Look Hear: Gaze Prediction for Speech-Directed Human Attention,rs, from 220 participants performing our referral task. In our quantitative and qualitative analyses, ART not only outperforms existing methods in scanpath prediction, but also appears to capture several human attention patterns, such as waiting, scanning, and verification. Code and dataset are avai