Train error about cornernet-lite HOT 9 OPEN

Not *.zip but *.jpg. And you also need create a folder named annotations which should contains instance_train/val{}.json for your training and testing in the coco dir.

from cornernet-lite.

the following is my structure according to the document
+CornerNet-Lite
+data
+coco
+images
+trainval2014
+ *.jpg
+minival2014
+ *.jpg
+testdev2017
+ *.jpg
+annotations
-instances_trainval2014.json
-instances_minival2014.json
-instances_testdev2017.json
but I met the same problem.

from cornernet-lite.

i met the same issue,how to solve it? thanks

from cornernet-lite.

The reason why you meet this issue is that your model can't read the image successful. So carefully check the xxx.py in core/dbs. The original repo has no bug with that problem and I also successfully train the CornerNet-Lite on my own dataset. The only drawback is that the training speed is so slow :<

from cornernet-lite.

i met the same issue,how to solve it? thanks

do you hava solve it? i met the same issue.my structure is like yours.

from cornernet-lite.

do you hava solve it?

from cornernet-lite.

I had the same problem, and found the workaround of it.

• trainval2014 <- jpgs from train2014.zip + val2014.zip
• minival2014 <- jpgs from val2014.zip
• testdev2017 <- jpgs from test2017.zip

from cornernet-lite.

Hi all, I met the same problem here. I only modified the gpu number and the rest of the codes are not touched.

It seems that the code is combining wrong pairs of directory and image image. For example, combining test directory with val image name, resulting in a wrong image file path.

from cornernet-lite.

I need help.I also meet this problem. Trying all the suggestions above,I can not get the true. ''NoneType' object has no attribute 'shape'' is being here whatever i do.

from cornernet-lite.