Test loss=0 with FIT_LARGE_SIZE_AND_PAD resize method

[Goldit]

Hi @weiliu89,

I am using FIT_LARGE_SIZE_AND_PAD resize_mode for training. The loss for training was fine, however, I got loss = 0 for testing.

I notice that there is a small different in resize_param for WARP and FIT_SMALL_SIZE.

'resize_param': { 'prob': 1, 'resize_mode': P.Resize.WARP, 'height': resize_height, 'width': resize_width, 'interp_mode': [P.Resize.LINEAR], },

resize_param = { 'prob': 1, 'resize_mode': P.Resize.FIT_SMALL_SIZE, 'height': resize_height, 'width': resize_width, 'height_scale': resize_height, 'width_scale': resize_height, 'interp_mode': [P.Resize.LINEAR], }

What would be the correct resize_param for training with FIT_LARGE_SIZE_AND_PAD resize mode?

Thank you and best regards, Nhan.

[weiliu89]

That part is not fully debugged. You could use ssd.ipynb and ssd_detect.ipynb to check if anything goes wrong.