- This app integrates Apple's two awesome frameworks CoreML, ARKit together.
- AR session gives CoreML input, and CoreML output tells ARKit where to place the virtual node for foot detected.
- Due to lack of data, the CoreML model might be biased. Works best with bare foot or wearing white socks on wooden floor.
- Semantic segmentation model trained on pytorch.
- Converts 224*224 image to 112*112 binary mask.
- Use U-Net as backbone.
- Use weights from MobileNet-v2 for transfer learning.
- Checkout code from this amazing repo.
- Upload a better version of the COVID-19 3D model. The one that's on the repo may be buggy.