Demo Application - TobiasSchmidtDE/DeepL-MedicalImaging GitHub Wiki
The demo application enables evaluating the predictions and visualizations of a single selected model or multiple models together.
Functionality
Sidebar
The sidebar allows the user to select whether to use a single model or an ensemble (predictions for the ensemble are averaged without weights) and to select the threshold for the bounding boxes. If the threshold is lower the bounding boxes and visualized areas will be larger and more general.
Main application
The user can select the model(s), that should be used for the prediction and then upload an image.
Once the image is selected the visualizations for the image with the selected models will be shown. Again the user can choose to customize the results by choosing the prediction threshold. All predictions under this threshold will the be ignored for the visualizations. The type of visualization, class-based or combined, can be chosen to either show separate visualizations for each class, or to get a combined image for all classes.
Using the Application
The application can either be used on http://131.159.94.14:8501/ (only accessible using the University VPN), locally by installing the dependencies and running the application, or using docker (described in the README)