Tag: #CLIP
Kubernetes, DevOps and MLOps, written from things we ran and measured ourselves · 2 posts
Contrastive Learning and CLIP: Learning by Pairing Images and Text to Guess Unseen Things
This explains how CLIP, trained solely on 400 million image-text pairs from the internet without human-assigned classification labels, solves previously unseen classification problems. We break down the InfoNCE loss appl
2026-10-17 · 14 min read #비전 모델#CLIP#대조 학습#InfoNCE#SigLIPOpen-Vocabulary Detection: Attaching CLIP Text Embeddings to Detectors — OWL-ViT, OWLv2, Grounding DINO, YOLO-World, YOLOE
Standard detectors only find 80 categories (COCO) fixed at training time. Given only the word "racket", an open-vocabulary detector starts by replacing the fixed weights of the classification head with embeddings generat
2026-10-17 · 15 min read #비전 모델#열린 어휘 검출#CLIP#OWL-ViT#OWLv2