Models¶
Symptoms Detection
zenkl_et_al_2025b(latest)zenkl_et_al_2025aanderegg_et_al_2024
Symptoms Segmentation
zenkl_et_al_2025b(latest) – uses 1/2 downscaling on the inputlatest_large– uses full resolutionzenkl_et_al_2025a– uses full resolutiontracking_latest– uses 1/2 downscaling on the inputanderegg_et_al_2024– uses full resolution
Organ Segmentation
zenkl_et_al_2025b(latest) – uses 1/4 downscaling on the input
Focus Estimation
Note
The
Depth Anything v2model is sensitive to input size due to TorchScript export constraints and cannot predict on arbitrary image sizes. Pull requests to fix this are welcome.Supported input sizes:
4096x6144(latest)6144x40964096x40962048x20481024x1024
Please note that the focus estimation uses 1/4 downscaling on the input.