search

hmsch / natural-synthetic-anomalies

Code for ECCV 2022 paper "Natural Synthetic Anomalies for Self-Supervised Anomaly Detection and Localization".
https://arxiv.org/abs/2109.15222
MIT License
48 stars 6 forks source link

about self_sup_tasks.py #3

Closed ghost closed 2 years ago

ghost commented 2 years ago

In self_sup_tasks.py, the way to create synthetic anomalies is different from the method described in the paper, for example, the patch width and height w h are selected according to formula 5 and formula 6, with gamma distribution (2, 0.1), but i didn't find anything like the formula in self_sup_tasks.py, actually formulas in section 3 are not clearly explained in the paper, what are Wmin and Wmax in formula 5? what is b in formula 8? in the paper it didn't define the symbols in the formulas, could you please explain how to generate synthetic anomalies?

ghost commented 2 years ago

I found them at the end of the paper, supplementary section, they are hyperparameters.