search

CharmonyShen / FireRisk

24 stars 1 forks source link

code #3

Open rongtongxueya opened 6 months ago

rongtongxueya commented 6 months ago

Great work. I was wondering if the code for your experiment is open? I used this dataset in a recent work, and I wanted to use the experiment in your article as a comparison experiment

sachith500 commented 2 months ago

Hi! Thanks for your interest.

Since the best results were from MAE ImageNet 1K, please refer to finetuning code in: https://github.com/facebookresearch/mae

If you want an alternate formulation of MAE with class balancing for unbalanced data etc try: https://github.com/sachith500/Sherlock (https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9889730) That work doesn't have an ImageNet baseline though...