Nguyen Van Cao

> Looks like this feature exists in the commercial product, any chance of it being included here? https://databricks.com/blog/2022/02/01/streamline-mlops-with-mlflow-model-registry-webhooks.html I don't think we can expect this feature in the main fork.

Just replace `icboard` by the keyword of the City Scape dataset, probably `cityscape`

> @bohare Hi were you able to test using test.py? Can you please share your test.py command? I am trying the following but it is not working: > `python test.py...

> @bohare @cao-nv Thanks for the reply. I am a little confused. You mentioned I need to point root to the Cityscapes data directory. So does FCHarDNet/test.py require the Cityscape...

Actually, the reported FPS was calculated with only the model time and post-processing time, not the entire pipeline.

This is because the path to images in your dataset, then `cv2` could not read them. You should: 1. Check path to the image folder in your `.py`. 2. Check...

ViT models are trained on Webface42M, and the authors confirmed in #1910 that they will not release models trained on that dataset.

> Then why do they put stats comparing Resnets and ViT ? Just for reference. Maybe due to some constraints so they cannot provide the trained models.

> what is the number of variables? pixel level? 224_224? 1024_1024 for HR img task? It seems like not a good idea, not the proper scenario, e2e using it directly,...

I evaluate the trained VGG16 model on Sdget task on Visual Genome and followings result: ``` R@20: 20.74 R@50: 29.36 R@100: 35.95 ``` These results are somewhat different from the...