Replies: 5 comments
-
any suggestion would be appreciated:) |
Beta Was this translation helpful? Give feedback.
-
You may train multiple times, and compare with https://download.openmmlab.com/mmpose/top_down/hrnet/hrnet_w32_crowdpose_256x192_20201227.log.json |
Beta Was this translation helpful? Give feedback.
-
We use 8 gpus. |
Beta Was this translation helpful? Give feedback.
-
Thanks so much for your reply. |
Beta Was this translation helpful? Give feedback.
-
That's interesting. I do not remember I changed the config files. Not sure why the learning rates are different :) |
Beta Was this translation helpful? Give feedback.
-
I followed the official config and trained HRNet-W32 on CrowdPose, however, I didn't get the results as good as the MMPose reported.
Mine(8*64) VS MMPose report : 66.2 AP VS 67.5 AP.
I was wondering that what are the official settings (e.g., how many GPUs used?)
Here comes the command I use:
tools/dist_train.sh configs/top_down/hrnet/crowdpose/hrnet_w32_crowdpose_256x192.py 8
The resulted log is as follows:
Beta Was this translation helpful? Give feedback.
All reactions