Question about libraries

[janlesniak-prog]

Hello,

For the semantic segmentation prediction in the Jupyter notebook, library dependencies seem to mismatch:

Requires mmseg >= 1.0, mmcv >= 2.0:
from mmseg.apis import init_model, inference_model, show_result_pyplot
from models.backbones import DinoVisionBackbone

Requires mmseg 0.27.0, mmcv == 1.5.0 (e.g. for mmcv.runner)
from dinov2.eval.segmentation.models.decode_heads.linear_head import BNHead

I was not able to arrive at one installation which allows to satisfy all imports. How could one resolve this dependency to make the code runnable? Thank you very much for this great example repository!

[zshn25]

Hi,

thank you for your interest. This repo uses mmseg >= 1.0. I had not faced any issues while using dinov2.eval.segmentation.models.decode_heads.linear_head's BNHead with mmseg>=1.0.

Please try with the proposed requirements and let me know if you get any error

[Brown-Ebouky]

Hi @janlesniak-prog ,

please did you solve this issue with the requirements? I am facing the same problem.
@zshn25 , which version of mmsegmentation are you using in your environment.

Thank you very much for your time.

[zshn25]

Please try with the following

conda create -n mmseg1.x python=3.9
conda activate mmseg1.x

# pip install -r ../dinov2/requirements.txt

pip install -U openmim
mim install mmengine
mim install "mmcv>=2.0.0"

git clone -b main https://github.com/open-mmlab/mmsegmentation.git
cd mmsegmentation
pip install -v -e .