DetCLIPv2: Scalable Open-Vocabulary Object Detection Pre-training via Word-Region Alignment.
Lewei YaoJianhua HanXiaodan LiangDan XuWei ZhangZhenguo LiHang XuPublished in: CoRR (2023)
Keyphrases
- object detection
- discriminatively trained
- keywords
- object detectors
- word alignment
- supervised learning
- face detection
- machine learning
- boosted classifiers
- computer vision
- co occurrence
- scene understanding
- training set
- training examples
- training process
- word level
- image classification
- region of interest
- input image
- out of vocabulary
- neural network