Zhu Zhen
Zhu Zhen
From your description, I'm sorry that I cannot figure out the reason. If you could, it would be better to experiment a little and see how it goes.
Our initial intention is trying to stabilize the training process and reduce artifacts. These techniques are of some help to these aspects but show no further potential to improve the...
From your provided information, I think most probably, the heat map of the pose that you fed is with the incorrect size. Please have a check.
Please see the [evaluation page](https://captain-whu.github.io/DOTA/evaluation.html)
Please follow the Installation instructions in README. Running init.sh or init.bat will automatically compile such library for you.
I think the problem should lie in data processing. Have a look. It might not be the format problem. Maybe some of your images are corrupted or the image path...
I think you should refer to [DOTA_devkit](https://github.com/CAPTAIN-WHU/DOTA_devkit). This should be a data processing problem. @dingjiansw101 can you help solve this?
Training on such a dataset with so many classes actually requires a huge budget of computation. It is because setting so many groups for a group conv would hugely decrease...
We cannot use the method mentioned in the SPADE paper to get the same results as the ones provided in their paper. It seems that we are not the only...