search

anilsathyan7 / Portrait-Segmentation

Real-time portrait segmentation for mobile devices
MIT License
638 stars 133 forks source link

PortraitNet Model #31

Closed shivSD closed 3 years ago

shivSD commented 3 years ago

@anilsathyan7 thanks for the great repo. Question PortraitNet model that you have used in this repo. Did you train the model from scratch or directly took the author's pre-trained model. I'm trying to train from the scratch so far unsuccessful, the training loss doesn't decrease after a few hundred epochs. Any suggestions on this would be useful.

anilsathyan7 commented 3 years ago

As mentioned in the section - Techniques For Improving Robustness, Accuracy and Speed, i trained the whole network from scratch using AISegment dataset. I had to modify the original code to train the model from scratch. It seems to use syntax/fucntions from an older version of tensorflow and pytorch.

Also make sure you use the correct dataset and training configuration, before training the model.