_______ ___ ___ ________ ___ ___ ___ ________ _________ ___ ________ ________
|\ ___ \ |\ \ / /||\ __ \ |\ \ |\ \|\ \ |\ __ \ |\___ ___\|\ \ |\ __ \ |\ ___ \
\ \ __/| \ \ \ / / /\ \ \|\ \\ \ \ \ \ \\\ \\ \ \|\ \\|___ \ \_|\ \ \\ \ \|\ \\ \ \\ \ \
\ \ \_|/__\ \ \/ / / \ \ __ \\ \ \ \ \ \\\ \\ \ __ \ \ \ \ \ \ \\ \ \\\ \\ \ \\ \ \
\ \ \_|\ \\ \ / / \ \ \ \ \\ \ \____ \ \ \\\ \\ \ \ \ \ \ \ \ \ \ \\ \ \\\ \\ \ \\ \ \
\ \_______\\ \__/ / \ \__\ \__\\ \_______\\ \_______\\ \__\ \__\ \ \__\ \ \__\\ \_______\\ \__\\ \__\
\|_______| \|__|/ \|__|\|__| \|_______| \|_______| \|__|\|__| \|__| \|__| \|_______| \|__| \|__|
self-defined useful comparasion functions, designed based on Kummerer, M., Wallis, T. S., & Bethge, M. (2018).
Saliency benchmarking made easy: Separating models, maps and metrics. ECCV (pp. 770-787).
the full code would consider image1 is the pred image, while the image2 is the ground truth image. Using carefully!
It is a non-official numpy&opencv implementation of Kummerer M, Wallis T S A, Bethge M. Saliency benchmarking made easy: Separating models, maps and metrics[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 770-787.
All code is documented in the evaluation.py file, containing NSS, IG, CC KL-Div & SIM for single image iteration.
Well annotated!