Not able to generate correct result

[imashu]

After running the predict_test.py from semantic segmentation code, I get a output of Validation data images in the /out directory. But to submit for evaluation scores on CodaLab we need results for the Tese data not the Val data. Please clarify what is wrong in this case.

This output is after running ./script/predict_test.sh. We can see clearly it is taking Val images as input to produce the ./out images
INFO:numexpr.utils:NumExpr defaulting to 2 threads.
Load model!
INFO:data.loveda:./LoveDA/Val/Urban/images_png -- Dataset images: 677
INFO:data.loveda:./LoveDA/Val/Rural/images_png -- Dataset images: 992
Downloading: "https://download.openmmlab.com/pretrain/third_party/hrnetv2_w32-dc9eeb4f.pth" to /root/.cache/torch/hub/checkpoints/hrnetv2_w32-dc9eeb4f.pth
100% 158M/158M [00:06<00:00, 25.6MB/s]
INFO:ever.core.logger:HRNetEncoder: pretrained = True
100% 417/417 [11:06<00:00, 1.60s/it]

[Junjue-Wang]

You can revise the config paths to generate test masks.