Add torchvision deeplabv3 Resnet50 and deeplabv3 mobilenet large #231
Conversation
oke-aditya
changed the title
|
✔️ Deploy Preview for pytorch-hub-preview ready! 🔨 Explore the source changes: efcc4e2 🔍 Inspect the deploy log: https://app.netlify.com/sites/pytorch-hub-preview/deploys/613653a3c31c2600071a59be 😎 Browse the preview: https://deploy-preview-231--pytorch-hub-preview.netlify.app |
| @@ -18,7 +18,10 @@ order: 1 | |||
|
|
|||
| ```python | |||
| import torch | |||
| model = torch.hub.load('pytorch/vision:v0.10.0', 'deeplabv3_resnet101', pretrained=True) | |||
| model = torch.hub.load('pytorch/vision:v0.10.0', 'deeplabv3_resnet50', pretrained=True) | |||
There was a problem hiding this comment.
With the change of model + the change of input image, would you mind verifying that the new output is still reasonable?
There was a problem hiding this comment.
I think it should be. As per the table there is little tradeoff
| Model structure | Mean IOU | Global Pixelwise Accuracy |
|---|---|---|
| deeplabv3_resnet50 | 66.4 | 92.4 |
| deeplabv3_resnet101 | 67.4 | 92.4 |
There was a problem hiding this comment.
I was more thinking in terms of output of the notebook, e.g. make sure that the segmentation of the new image isn't completely off
There was a problem hiding this comment.
Yeah, I will have a run locally and post outputs of both the models.
Edit: all three models in this case.
There was a problem hiding this comment.
Did you have a chance to check the segmentation output @oke-aditya ?
Otherwise no worries, I'll check myself just let me know :)
There was a problem hiding this comment.
Sorry for the delay! (I was away for the week) I will post them in few hours today.
There was a problem hiding this comment.
Mobilenetv3
Resnet50
Resnet101
I think resnet50 can be a good tradeoff. It took significantly lower time to process over my CPU machine.
There was a problem hiding this comment.
Agreed! Thanks a lot of the outputs
|
Sorry for all the delays. From this weekend I will have more time to work on Pending PRs |
No worries at all @oke-aditya , there's no real rush :) Let's merge this one, thanks a lot for your PRs! |
Helps pytorch/vision#4289
Adds the following models.
Edit: -
I checked the netlify preview. It was as expected.
Feel free to merge.
cc @NicolasHug @datumbox