results of voc-15-5 - Githubissues

arthurdouillard / CVPR2021_PLOP

Official code of CVPR 2021's PLOP: Learning without Forgetting for Continual Semantic Segmentation

https://arxiv.org/abs/2011.11390

MIT License

143 stars 23 forks source link

results of voc-15-5 #19

Closed Richardych closed 2 years ago

Richardych commented 2 years ago

Hi,

I run your code on voc-15-5, and get: Final Mean IoU 69.35 Average Mean IoU 74.84 Mean IoU first 75.25 Mean IoU last 50.47

however, the paper stated is: 75.73 51.71 70.09 75.19

why the results are different and poor than the paper stated?

arthurdouillard commented 2 years ago

Paper vs your run:

75.73 vs 75.25
51.71 vs 50.47
69.35 vs 70.09
74.84 vs 75.19

The difference is always small (except maybe for old). It's probably due to the inherent randomness of the initialization (can change depending on cuda version, pytorch version, etc.).

I don't see that as a major difference. Have you look at 15-1 ?