search

yandex-research / ddpm-segmentation

Label-Efficient Semantic Segmentation with Diffusion Models (ICLR'2022)
https://yandex-research.github.io/ddpm-segmentation/
MIT License
662 stars 60 forks source link

Need more details reproducing Table 2 results #34

Closed srk1995 closed 11 months ago

srk1995 commented 1 year ago

Hi, I'm trying to reproduce Table 2 in your paper, but I got a 50.59% mIoU in FFHQ-34, while Table 2 shows 59.1%. Additionally, I obtained a 30.87% mIoU in Bedroom-28, but Table 2 shows 49.4%. Could you please provide more information to help me reproduce your results?

I used an uploaded config file and my system configuration is as follows:

Operating System: Ubuntu 22.04 Python version: 3.7.16 GPU: TITAN RTX Seed: 0 (default in your code)

Thank you.

dbaranchuk commented 1 year ago

Hi, this instruction has to lead to almost perfect reproducibility: https://github.com/yandex-research/ddpm-segmentation#run. Please also consider setting up your environment according to https://github.com/yandex-research/ddpm-segmentation/blob/master/requirements.txt.

We also provide the pretrained checkpoints here, so first you can try to evaluate them.