Message type "caffe.LayerParameter" has no field named "black_augmentation_param"

[yxqlwl]

Hi, Sorry for troubling. Errors occurred in the training process:

1. "Message type "caffe.LayerParameter" has no field named "black_augmentation_param"" I cannot find the definition of black_augmentation in src/caffe/layers or caffe.proto. I am wondering if I missed something?
   2. Unknown bottom blob 'img0_a_org' (layer 'crop_params') Thanks~

[brjeon]

I believe img0_a_org stands for the original Image 0. Therefore, I have changed some names in the train.prototxt as following. blob0 -> img0_a_org blob1 -> img1_a_org blob2 -> flow_gt_a_org Correct me if I am wrong.

However, I still cannot figure out what img0_b, img1_b, flow_gt_b stand for.

[yxqlwl]

Thanks, brjeon. I also think so. I will check it. But as for the black_augmentation_param, I think sth. is missing.

[nikolausmayer]

Hi yxqlwl, I cannot find any mention of a black_augmentation_param in the codebase. I'm assuming that the error occurs in a layer you added?

[brjeon]

Hi nikolausmayer, For your information the 'black_augmentation_param' is mentioned in 'FlowNet2-CSS-ft-sd_train.prototxt.template'.

[yxqlwl]

Hi nikolausmayer, As brjeon mentioned, it is found in 'FlowNet2-CSS-ft-sd_train.prototxt.template'. May be there are different versions of the code? :)

[brjeon]

Actually, the above mentioned 'FlowNet2-CSS-ft-sd_train.prototxt.template' file is not in the github codebase. It is created after running 'download-models.sh' file which is located in the 'flownet2-master/models' directory.

[nikolausmayer]

Woops sorry, now I see. I didn't check the models. We'll look into what's missing.

[yxqlwl]

Thanks~

[eddy-ilg]

Thanks for pointing this out, the code part it missing in the repo. I will add it during the week.

[eddy-ilg]

Black augmentation layer is now in the repo.

[yxqlwl]

Thank you!