Closed ghost closed 11 months ago
We have just checked the checkpoint, and everything is fine. Can you please share the full log of your test run?
Sorry, I was able to reproduce your result by setting all the versions of libraries exactly the same as your Dockerfile. Before, I was using the lower version of pytorch. Closing now.
@dev0812 Can you share the environment.yaml file. If you used conda ?
@jaswanthbjk if it helps, i can share the ouput of conda env export
from our docker image:
First of all, Thank you very much for your great work!
When I ran your code, the evaluation result was far worse than what you reported. I am getting 74.0, 66.8, 49.3 for AP_0.25, AP_0.50 and AP from your test script, whereas I am supposed to get 86.7, 78.3, 59.3 for AP_0.25, AP_0.50 and AP.
The result in AP is 12.7 lower than your reported result, which is unreasonable. I am using the validation checkpoint(oneformer3d_1xb4_scannet.pth) and config(oneformer3d_1xb4_scannet.py) that you provided for ScanNetV2 validation set result.
Could you let me know how to fix it?