cbsong
cbsong
> 就是把不同scale的输出合并了，然后把最后两个维度交换了一下。2100=80x80+40x40+20x20,也就是说总共有2100个输出，112=80+4x8，每个输出有80类的score，和4个框位置信息，每个框位置信息是由8个值构成的分布。需要自己做后处理，得出实际有效的检测框。 能否写个python示例~
Thanks ! The stride of conv is 2, and the input size is 28x28.
图片404了
> We will publish the training code for [ALIKED](https://github.com/Shiaoming/ALIKED) (the subsequent work of ALIKE, which we are working on). Then you can refer to the training code of ALIKED, they...
Hi, could you share the training code with me as well? [email protected] Thank you in advance :)
``` [{'from': 'human', 'value': ' What can you see in the following the second image'}, {'from': 'gpt', 'value': 'The image is a photograph of a two-story residential house during the...
Sorry, I wonder when to support the Pixtral.