Описание
MWebMLPredict is a cutting edge machine learning tool that can be used by Radiologists to predict various diseases from a given Radiology image (in JPEG, PNG or DICOM formats). The ML Inference computation happens on the device as the trained model is downloaded to the device and used to do the prediction. Images that are available on the Camera roll can be used to check for various diseases. Currently, one model for Breast imaging is available. The mode is hosted on the server and is downloaded to the mobile app's browser engine for execution. This model is based on Tensorflow.js. As new disease prediction models are made available, they will be available in the mobile app without the need to update the mobile app as the new models are downloaded from our website. The integrated camera function enables Radiologist to take images of Radiology posters with images and then use them to predict various diseases. The images used for prediction never leave the device protecting the privacy of the user. The application doesn't transmit any image back to the server. It only retrieves latest trained model from server when the application is started.