search

rishizek / tensorflow-deeplab-v3-plus

DeepLabv3+ built in TensorFlow
MIT License
833 stars 307 forks source link

Training in two classes #46

Open sonfire186 opened 5 years ago

sonfire186 commented 5 years ago

I need to train the network only in two classes from PASLAC VOC. Background and personal.

train.py

_NUM_CLASSES = 2

preprocessing.py

label_colours = [(0, 0, 0),  # 0=background
                 (192, 128, 128) # 15=person]

Am I doing the right thing?

northeastsquare commented 5 years ago

Yes, if your training set, person label, is 192,128,128, that should work. I use (128,0,0) to annotate the other class

DomicZhong commented 4 years ago

what should I set If I want to retrain the model in two classes， that is， I want to the model output two classes