search

EnVision-Research / Generalizable-BEV

147 stars 18 forks source link

Seeking Clarification on Missing Files and Training Results #5

Closed shhwang0129 closed 2 months ago

shhwang0129 commented 3 months ago

Thank you for the fantastic work! I'm very interested in your approach.

While trying to run the code, I encountered a few issues and would appreciate your guidance on them.

  1. pretrained backbone (resnet50-0676ba61.pth) I noticed that this specific pretrained backbone is not included in the repository. I replaced it with 'torchvision://resnet50'.

  2. 'lyft_infos_ori_ann_train.pkl' Following the README instructions for preparing the Lyft data, I was only able to obtain 'lyft_infos_ann_train.pkl'. I used this file as a replacement.

Despite making these adjustments, the results I obtained seem to differ from those reported. Could you provide insights into what might be causing these discrepancies?

Thank you in advance for your assistance!

LuPaoPao commented 3 months ago

Thank you for the fantastic work! I'm very interested in your approach.

While trying to run the code, I encountered a few issues and would appreciate your guidance on them.

1. pretrained backbone (resnet50-0676ba61.pth)
   I noticed that this specific pretrained backbone is not included in the repository. I replaced it with 'torchvision://resnet50'.

2. 'lyft_infos_ori_ann_train.pkl'
   Following the README instructions for preparing the Lyft data, I was only able to obtain 'lyft_infos_ann_train.pkl'. I used this file as a replacement.

Despite making these adjustments, the results I obtained seem to differ from those reported. Could you provide insights into what might be causing these discrepancies?

Thank you in advance for your assistance!

Thank you for your attention. What is the specific result? And what kind of measurements are you doing from that dataset to that dataset? If you can give me specific indicators, I will give you more answers .Because I'm not sure what's wrong with you. I would appreciate it if you could provide as many results as possible.