search

hellochick / PSPNet-tensorflow

TensorFlow-based implementation of "Pyramid Scene Parsing Network".
326 stars 123 forks source link

ignoring label 0 in ade20k doesn't give the same labeling as the check point results #50

Closed AyaMohamedS closed 5 years ago

AyaMohamedS commented 6 years ago

i set index 0 to be ignored for ade20k training but in testing my trained model, i got class label 0 and person's label becomes 13 not 12 as in the check point model what can i do to fix this problem? and what is the expected labeling output for a model trained on the ade20k_2016challenge?

AyaMohamedS commented 5 years ago

i solved the problem by adjusting image reader (random_crop_and_pad_image_and_labels function) to convert label 0 into 255 before subtracting 1 from label tensor then convert 254 to 255

and i kept ignore label =255 in train.py

tjqinwind commented 5 years ago

can I have a look at you code? i still have a litter confuse.