search

zhangy0822 / USER

USER: Unified Semantic Enhancement with Momentum Contrast for Image-Text Retrieval, TIP 2024
20 stars 0 forks source link

代码中未找到CLIP-Guided Enhancement (CGE) module #6

Closed XIANGLIU03 closed 1 week ago

XIANGLIU03 commented 4 months ago

你好，你们的工作非常棒！在代码在未找到使用CLIP visual encoder的代码，请问要如何使用CGE模块？

zhangy0822 commented 4 months ago

Thanks for your attention! Please refer to the issue https://github.com/zhangy0822/USER/issues/4#issuecomment-2014270919.

zc020126 commented 2 months ago

你好，我尝试发现您的代码，但是跑完f30k的数据集后，结果如下： 2024-08-13 03:06:19,855 calculate similarity time: 0.03725838661193848 2024-08-13 03:06:20,003 Image to text: 0.1, 0.3, 0.4, 2491.0, 2482.1 2024-08-13 03:06:20,078 Text to image: 0.1, 0.5, 1.0, 500.0, 500.5 2024-08-13 03:06:20,078 Current rsum is 2.4 请问可能出现的原因是什么呢？

zhangy0822 commented 1 month ago

你好，我尝试发现您的代码，但是跑完f30k的数据集后，结果如下： 2024-08-13 03:06:19,855 calculate similarity time: 0.03725838661193848 2024-08-13 03:06:20,003 Image to text: 0.1, 0.3, 0.4, 2491.0, 2482.1 2024-08-13 03:06:20,078 Text to image: 0.1, 0.5, 1.0, 500.0, 500.5 2024-08-13 03:06:20,078 Current rsum is 2.4 请问可能出现的原因是什么呢？

Please provide the complete training script. In addition, has the loss decreased?