📍 GEM: Unlocking the Latent Localization Ability of VLMs! Title: Grounding Everything: Emerging Localization Properties in Vision-Language Transformers Conference: CVPR 2024 Code/Checkpoi...
개요 이 포스트에서는 VL-SAM의 핵심인 “객체 이름 → 위치 힌트(Attention Map) 생성 → SAM 포인트 프롬프트” 중, Attention Map 생성(VLM 측) 실습을 다룹니다. (반복 iteration 없이, 단일 패스 성격의 데모) 1) [Object Recognition] – Attention Map Generation ...
🔎 VL-SAM: Training-Free Open-Ended Object Detection & Segmentation Title: Training-Free Open-Ended Object Detection and Segmentation via Attention as Prompts Conference: NeurIPS 2024 ...
🔎 (English) CLIP Surgery: Enhancing Explainability by Operating on CLIP! Title: A Closer Look at the Explainability of Contrastive Language-Image Pre-training (CLIP Surgery) Journal: Patter...
🧩 (한국어) PartCLIPSeg: CLIP으로 파트 단위까지 인식하는 Open-Vocabulary 세분화! 제목: PartCLIPSeg: Understanding Multi-Granularity for Open-Vocabulary Part Segmentation 학회: NeurIPS 2024 코드/체크포인트: GitHub – Pa...
🧬 (English) Practicing ClipSurgery!! Today we will practice with the CLIP model that has been surgically modified — ClipSurgery! We’ll run it directly and generate a clean similarity map (CAM) on...
🧬 (English) EfficientSAM Practice!! Today, let’s revisit the practice of the new SAM model SAM2, which we studied theoretically in a previous post and also experimented with before! This time, in...
🔎 Open-Vocabulary SAM: Expanding to Recognize & Segment 20,000 Classes! Title: Open-Vocabulary SAM: Segment and Recognize Twenty-thousand Classes Interactively Conference: ECCV 2024 C...
🧩 (한국어) SAMEO : 가려진 객체까지 한 번에 segmentation!! 제목: Segment Anything, Even Occluded (SAMEO) 학회: CVPR 2025 프로젝트/데모: Project Page · CVF OpenAccess PDF 핵심 키워드: Amodal Instance Segmentation, S...
🧬 (English) EfficientSAM Practice!! Today, following the previous post where we studied the theory, we will conduct a hands-on experiment with the lightweight segmentation model EfficientSAM! ...