There are several options in the Settings section to experiment with (please see examples below) where you can select different styles (more on the way!), completely replace the background of your picture, make the background blurred out, etc.
Note: Since these models run locally on device, any uploaded pictures stay safely on the device. Please note that the first run may be much slower than subsequent runs as Tensorflow loads the necessary files. (WIP so there may be bugs!)
|
Output type: |
Style: |
|
Style quality: |
|
|
Blur background: |
Use a different background: |
|
Select an image: |
Upload an image: |
Enter URL: |
|
Upload an image: |
Enter URL: |
This demo combines style transfer models (please see the details section of my Real-time style transfer demo for information on how the models were optimised using techniques such as pruning) and a semantic segmentation model to extract people out of an image. The semantic segmentation model is based on MobileNetV3 and was mostly trained on open datasets and lightly pruned to give the final model.
All models were trained and optimised in Pytorch and then converted for deployment on TensorFlow.js. For deployment, all the layers in the models that could be fused were fused and the models quantised to give the final optimised models.
This demo was put together by Zawar Qureshi but could not have been done without the following:
I'd love to hear from people interested in making tools/apps using these models!