Colon Polyp Recogn

Features Colon Polyp Recogn

Prototype application for the recognition of polyps or healthy tissue from colonoscopy image clippings.
Image recognition is performed with a pre-trained neural network using TensorFlow and TensorFlow Lite.
This network has been trained with clippings from the CVC-ClinicDB database.
The predictor model is limited to polyp and non-polyp clippings only.
In validation, an accuracy of 0.92 is achieved with the non-polyp class and 0.98 with the polyp class.
The app is ad-free and does not collect any personal information.
In the future it is expected to improve the current model and implement new models with new algorithms such as SSD MobileNet and YOLO.
Made and designed by the RNASA-IMEDIR laboratory, CITIC, University of A Coruña (UDC).
If you have any suggestions, please contact us at [email protected].