bityangke / Weakly_supervised_fire_segmentation

Weakly supervised fire segmentation with mid-layer visualization

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Weakly supervised fire segmentation with intermediate layer visualization
The code corrosponds to the paper "Weakly-supervised fire segmentation by visualizing intermediate CNN layers"

It contains two networks to be trained:
train2.py is used to train the first network to obtain initial visulization masks
train.py trains a deeplabv3 pixel-supervised segmentation CNN with the intial masks

To Evalue the performance on the test set run: evaluate.py for evaluating initial masks, and evaluate_segment.py for evaluating the second network and the final performance

In order to get output of image examples run: test.py for intial masks and test_segment.py for the final network

About

Weakly supervised fire segmentation with mid-layer visualization


Languages

Language:Python 100.0%