• Pixel Accuracy
• mean Accuracy(of per-class pixel accuracy)
• mean IOU(of per-class Mean IOU)
• Frequency weighted IOU

For more information, kindly refer Fully Convolutional Networks for Semantic Segmentation

Use

python demo.py convert_prediction predict_loc id_file

Use

python demo.py convert_gt gt_loc id_file

Use

python demo.py find_metrics predict_path gt_path id_file [--options]

• utils.py: contains functions.
• demo.py: contains a brief demo of how to use the functions.[use demo.py -h]

A few parts have been adopted from the code present in martinkersner/py_img_seg_eval. Although the formulations are slightly wrong, it was very helpful.

Also, a big thanks to Video Analytics Lab.