FocusCLIP: Multimodal Subject-Level Guidance for Zero-Shot Transfer in Human-Centric Tasks
收藏DataCite Commons2024-12-16 更新2025-04-16 收录
下载链接:
https://service.tib.eu/ldmservice/dataset/ca037458-1a4e-402d-ba71-fff0c65e61a0
下载链接
链接失效反馈官方服务:
资源简介:
FocusCLIP: Multimodal Subject-Level Guidance for Zero-Shot Transfer in Human-Centric Tasks. This paper introduces FocusCLIP, an enhancement for CLIP pretraining using a new ROI encoder. This encoder uses heatmaps to help the model focus on key image areas, improving performance.
提供机构:
TIB
创建时间:
2024-12-16



