sayzardSee-through은 단일 애니메이션 정지 이미지를 최대 23개의 의미적·완전 인페인팅된 레이어(머리, 얼굴, 눈, 의상 등)와 추정 드로잉 순서로 자동 분해해 계층화된 PSD와 깊이·마스크 인터미디어트를 출력하는 오픈소스 프레임워크입니다. LayerDiff, Marigold(의사깊이), SAM 등 모델을 통합하며 UI·데모·스크립트 제공. SIGGRAPH 2026 조건부 채택. Image→Live2D 전체는 아니지만 편집·리깅 전 단계 자동화에 유용합니다.
Chinstrap CommunityLightly - Computer vision suite for AI optimization
Cossmology Profile: https://dub.sh/b2oqpNO
Key People: Igor Susmelj, Matthias Heller
Phil NelsonI'm organizing a conference again this year! The OpenCV-SID Conference on Computer Vision & AI is this May 4th in Los Angeles. I have speakers from Disney, Bonsai Robotics, Code19 Racing, Ultralytics, Procter & Gamble, and Looking Glass Factory!
+ access to the Display Week exhibition hall, the largest gathering of display technology in the world. It'll be the most fun you're legally allowed to have at a tech conference.
OpenCV-SID Conference On Computer Vision & AI
OpenCV is organizing the 2nd annual OpenCV-SID Conference on Computer Vision and AI (OSCCA) to be held on May 4th 2026 in Los Angeles, and in conjunction with Display Week, the premier gathering of display technology professionals. Learn more & Register Talks & Speakers We have curated an awesome slate of speakers from around the […]
AA (@measure_plan)
three.js, tsl-webgpu, mediapipe 손 추적, kimi k2.5를 활용해 라센간(Rasengan) 블라스트 효과를 구현한 데모와 코드 공유 트윗입니다. AI/컴퓨터비전과 웹GPU를 결합한 창의적 인터랙티브 시연으로, 개발자들에게 참고할 만한 흥미로운 구현 사례입니다.
hesselschutde-embedding
Amsterdam, March 2026
Activations in the first layer of the YOLO (You Only Look Once) object detection AI model determine from which video frame to load pixels.
#urbanphotography #computationalart #generativeart #yolo #computervision #abstractstreet #amsterdam #video
Amsterdam, March 2026
Activations in the first layer of the YOLO (You Only Look Once) object detection AI model determine from which video frame to load pixels.
#urbanphotography #computationalart #generativeart #yolo #computervision #abstractstreet #amsterdam #video
Harald KlinkeVision-language models are impressive—until you ask them something simple.
A recent study shows that state-of-the-art systems struggle with basic visual tasks like counting shapes or detecting overlaps, achieving only ~58% accuracy on average—far below human performance
So what are they actually “seeing”?
AI doesn’t perceive images the way we do. It approximates, infers, guesses. And sometimes, it fails where humans succeed instantly.
#AI #ComputerVision #AIRealityCheck
Vision Language Models Are Blind | Anh Totti Nguyen
While large language models with vision capabilities (VLMs), e.g., GPT-4o and Gemini-1.5 Pro, are powering various image-text applications and scoring high on many vision-understanding benchmarks, we find that they are surprisingly still struggling with low-level vision tasks that are easy to humans. Specifically, on BlindTest, our suite of 7 very simple tasks such as identifying
AA (@measure_plan)
얼굴 사진과 링크드인용 프로필 이미지를 무료로 생성하는 앱 ‘Cursed Potato Head’가 공개되었다. 최신 컴퓨터 비전과 얼굴 인식 기술을 활용해 셀피를 전문적인 헤드샷으로 바꾸는 소비자용 AI 응용 사례로 주목된다.
Muhammad Rizwan Munawar (@muhammdrizwanmr)
공항에서 수천 개의 짐을 실시간으로 세어 운영 효율을 높이는 시스템을 직접 구축했다. 수하물 추적을 수동으로 처리하던 비효율을 개선하기 위해 Ultralytics의 YOLO26을 활용한 실시간 luggage counting AI를 구현한 사례다.
https://x.com/muhammdrizwanmr/status/2038101917466063142
#yolo #ultralytics #computervision #airportoperations #aiapplication
Muhammad Rizwan Munawar (@muhammdrizwanmr) on X
Real-time luggage counting at airports for smoother operations 🛄✈️ Airports handle thousands of suitcases and carry-ons every day. Tracking each bag manually is slow and error-prone, so I automated it. I built a real-time luggage counting system using @ultralytics YOLO26.
fly51fly (@fly51fly)
통합 토큰화와 잠재 노이즈 제거를 위한 엔드투엔드 학습을 제안한 최신 연구입니다. CV 분야에서 토큰화와 denoising을 함께 최적화하는 접근으로, 새로운 모델 학습 방법론에 관한 기술적 발전을 다룹니다.
https://x.com/fly51fly/status/2038028760512139761
#computervision #tokenization #denoising #training #research
Andreas BeckerAlexander Dobrindt und das Bundesinnenministerium forcieren den Einsatz von Edge-KI zur Echtzeit-Gesichtserkennung an Bahnhöfen.
Computer-Vision-Modelle analysieren Videostreams lokal auf verdächtige Bewegungsmuster und scannen auf Waffen, während biometrische Daten kontinuierlich mit Polizeiregistern abgeglichen werden.
#ComputerVision #EdgeAI #Biometrie #Datenschutz #News
https://www.all-ai.de/news/news26/bahnhof-ueberwachung-ki
