Whether there is an error in param_loss ?

[liqidong]

https://github.com/choyingw/SynergyNet/blob/072c577ec2add613f87c81cfc882c72c9a9a87c6/loss_definition.py#L40

[liqidong]

I think the id is not match. Whether I understand correctly or not？12:40:10 （pose:shape:exp）

[choyingw]

yes, its pose, shape, expr for 12:40:10

[ken881015]

@liqidong I think it's weird, too.

In "normal" mode, loss was defined below https://github.com/choyingw/SynergyNet/blob/072c577ec2add613f87c81cfc882c72c9a9a87c6/loss_definition.py#L37

So, It reasonable to think that the order of pose, shape, exp output by PointNet(reverse direction) is same as ground truth.

https://github.com/choyingw/SynergyNet/blob/072c577ec2add613f87c81cfc882c72c9a9a87c6/loss_definition.py#L40 In my opinion, the index of line 40 should be altered to

loss = self.criterion(input[:,12:62], target[:,12:62]).mean(1)

[choyingw]

Yes sorry it was a typo when I was reorganizing my code before. Please correct line 40 to 12:62 not :50

Thanks!

On Tuesday, June 20, 2023, ken881015 @.***> wrote:

@liqidong https://urldefense.com/v3/__https://github.com/liqidong__;!!LIr3w8kk_Xxm!qQlYFSOAnfWfq1H18unyMXWUCfKkbV7h9A8Pf7wHAUryILSQUNr8X5uNbMke3u7MVSBx5h95m-I5P-fnPSMwJ8Be$ I think it's weird, too.

In "normal" mode, loss was defined below https://github.com/choyingw/SynergyNet/blob/072c577ec2add613f87c81cfc882c7 2c9a9a87c6/loss_definition.py#L37 https://urldefense.com/v3/__https://github.com/choyingw/SynergyNet/blob/072c577ec2add613f87c81cfc882c72c9a9a87c6/loss_definition.py*L37__;Iw!!LIr3w8kk_Xxm!qQlYFSOAnfWfq1H18unyMXWUCfKkbV7h9A8Pf7wHAUryILSQUNr8X5uNbMke3u7MVSBx5h95m-I5P-fnPfcc93Xq$

So, It reasonable to think that the order of pose, shape, exp output by PointNet(reverse direction) is same as ground truth.

https://github.com/choyingw/SynergyNet/blob/072c577ec2add613f87c81cfc882c7 2c9a9a87c6/loss_definition.py#L40 https://urldefense.com/v3/__https://github.com/choyingw/SynergyNet/blob/072c577ec2add613f87c81cfc882c72c9a9a87c6/loss_definition.py*L40__;Iw!!LIr3w8kk_Xxm!qQlYFSOAnfWfq1H18unyMXWUCfKkbV7h9A8Pf7wHAUryILSQUNr8X5uNbMke3u7MVSBx5h95m-I5P-fnPQcNV1_g$ In my opinion, the index of line 40 should be altered to

loss = self.criterion(input[:,12:62], target[:,12:62]).mean(1)

— Reply to this email directly, view it on GitHub https://urldefense.com/v3/__https://github.com/choyingw/SynergyNet/issues/18*issuecomment-1600030352__;Iw!!LIr3w8kk_Xxm!qQlYFSOAnfWfq1H18unyMXWUCfKkbV7h9A8Pf7wHAUryILSQUNr8X5uNbMke3u7MVSBx5h95m-I5P-fnPSoqS-zH$, or unsubscribe https://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AKMLAJLRNJLOVWQD3LVTIOLXMJU37ANCNFSM5T67VMIQ__;!!LIr3w8kk_Xxm!qQlYFSOAnfWfq1H18unyMXWUCfKkbV7h9A8Pf7wHAUryILSQUNr8X5uNbMke3u7MVSBx5h95m-I5P-fnPTAtlKeW$ . You are receiving this because you commented.Message ID: @.***>

-- Ph.D. candidate of Computer Science Department University of Southern California

[ken881015]

@choyingw I think that I found another typo in released code. But I still want to discuss with you.

If you want to reverse the order of y axis (from 0~199 to 199~0), then code should be minus 1 instead of plus 1, do I think correctly?

[choyingw]

Ohhh yes it should be -1 not +1. Thanks for pointing out.