Thanks a lot for sharing the code! I have some questions about part segment.

[xindeng98]

Are the best parameters is the default parameters?Also I noticed that in the classification you use the default initialization, but in this task you use the xavier initialization method, is there any intuitive explanation for this? I hope to get your reply, maybe the parameter settings. Thank you very much for your contribution!

[ma-xu]

@xindeng98

1. yes.
2. I just followed previous works like GDANet for part segmentation task.
   Please let me know if you have any further questions.

[xindeng98]

@xindeng98

1. yes.
2. I just followed previous works like GDANet for part segmentation task.
   Please let me know if you have any further questions.

Thank you for such a quick reply! Can I get the training log of the part segment from you? I want to quickly judge whether my idea is feasible according to the comparison of loss descent speed.

[ma-xu]

@xindeng98 Sure. You can check commit:d2b8dbaa. The log files are provided here: [anonymous google drive]

[xindeng98]

Thanks a lot.