924973292 / TOP-ReID

【AAAI2024】TOP-ReID: Multi-spectral Object Re-Identification with Token Permutation
MIT License
42 stars 2 forks source link

该项目支持market1501训练吗 #8

Closed starsky68 closed 3 months ago

starsky68 commented 3 months ago

该项目支持market1501训练吗?有相应的参数吗

924973292 commented 3 months ago

TOP-ReID is consistent with EDITOR; you can simulate the input of Market1501 as tri-modal. #5
1.You can just change the dataset name in the config file. 2.You should adjust the following code to simulate tri-modal input:

The single-modal input format:

截屏2024-06-04 11 12 24

The multi-modal input format:

截屏2024-06-04 11 14 06

For example, you can copy the RGB image path to the NIR/TIR path. Ensure that your format matches the code below:

img.append(img_path_RGB) img.append(img_path_RGB) img.append(img_path_RGB)

starsky68 commented 3 months ago

TOP-ReID is consistent with EDITOR; you can simulate the input of Market1501 as tri-modal. #5 1.You can just change the dataset name in the config file. 2.You should adjust the following code to simulate tri-modal input:

The single-modal input format: 截屏2024-06-04 11 12 24

The multi-modal input format: 截屏2024-06-04 11 14 06

For example, you can copy the RGB image path to the NIR/TIR path. Ensure that your format matches the code below:

img.append(img_path_RGB) img.append(img_path_RGB) img.append(img_path_RGB)

感谢回复,请问这个多卡训练支持吗,如何设置

924973292 commented 3 months ago
截屏2024-06-04 11 32 01

Yes, the multi-GPU training setup is consistent with TransReID. You can adjust the parameters mentioned above in the config file for training. For details, you can refer to the relevant descriptions in the TransReID repository.