recognize-anything icon indicating copy to clipboard operation
recognize-anything copied to clipboard

Published 20 hours ago • verified publisher icon xinyu1205

how can i get det location

Open feifaxiaoming opened this issue 1 year ago • 6 comments

我在使用inference_ram.py 进行识别的时候，我如何在获取到标签的同时，能把标签对应的位置信息获取到呢。

feifaxiaoming avatar Jun 15 '23 09:06 feifaxiaoming

Please refer to RAM/Tag2Text with Grounded-SAM. RAM/Tag2Text provides image tags, while Grounded-SAM generates corresponding bounding boxes and masks.

xinyu1205 avatar Jun 15 '23 09:06 xinyu1205

您这个就是单独用于识别的是吧，但是您这个识别之前，也需要检测啊 ，检测的位置信息，不在这个代码中体现吗？

feifaxiaoming avatar Jun 15 '23 09:06 feifaxiaoming

识别前不需要检测，直接对一张图像输出标签

xinyu1205 avatar Jun 15 '23 09:06 xinyu1205

还是有点没懂，识别前不需要检测，那检测完，怎么把标签挂到对应的位置上呢

feifaxiaoming avatar Jun 15 '23 09:06 feifaxiaoming

通过grounding dino来根据tag生成bounding box

xinyu1205 avatar Jun 15 '23 10:06 xinyu1205

明白了，谢谢

feifaxiaoming avatar Jun 15 '23 10:06 feifaxiaoming