MSeg, universal demo takes 10 minute for 1 image, why so?

[luda1013]

Hallo Pros,

i am currently working with enhancing image enhancement paper and algorithm and trying to implement that. In the process, we need to use MSeg-segmentation for real and rendered images/ datasets. i have like 50-60k images.

So the dependencies MSeg-api and MSeg_semantic were already installed. I tried the google collab first and then copying the commands, so i could run the script in my linux also. the command is like this:
python -u mseg_semantic/tool/universal_demo.py
--config="default_config_360.yaml"
model_name mseg-3m
model_path mseg-3m.pth
input_file /home/luda1013/PfD/image/try_images

the weight i used, i downloaded it from the google collab, so the mseg-3m-1080.pth

but for me, it took like 10 minutes for 1 image and also what i get in temp_files is just the gray scale image of it.
Could someone help me how i could solve this problem, thank you :)

[luda1013]

My setup:

• Ubuntu 20.04.6 LTS
• core i9-10980XE @ 3GHz
• graphic (nvidia-smi): NVIDIA RTX A5000 (48 GB)
• Cuda 11.7
• Pytorch version : 2.0.0
• Cuda at: /usr/local/cuda*

[mseg-dataset]

Hi @luda1013, thanks for your interest in our MSeg models. It sounds like your GPU is not being utilized or recognized in your Pytorch installation. Can you please verify the following:

• Can you run nvidia-smi during inference to confirm that the GPU is actually being detected by Pytorch? Several GB of GPU RAM should be shown as being utilized when you start inference.

• Have you tried running one of your images in our Colab listed on our readme? There, we run the mseg-3m-1080p.pth model with
  config "mseg-semantic/mseg_semantic/config/test/default_config_360_ms.yaml", from https://github.com/mseg-dataset/mseg-semantic/releases/download/v0.1/mseg-3m-1080p.pth, and inference takes less than 10 sec on each of the demo images, even on an old Tesla T4 GPU provided for free on Colab.

• What resolution are your input images? The demo images in the colab have resolution (1080, 1920, 3), (1080, 1728, 3),(1080, 1920, 3), and (500, 990, 3) pixels.