Cancer Detection — Image Classifier
A model trained to sort an image into three classes: normal, benign, malignant.
The starting point.
An image-classification model trained with Google Teachable Machine, with no external code. It takes a 224 × 224 pixel image and outputs a probability for each of its three classes: normal, benign and malignant. The scores are raw probabilities with no decision threshold, so no result is ever turned into a recommendation. The model runs entirely in the browser with TensorFlow.js, so no image leaves the machine. It is an educational demonstration: it is not clinicallyically validated, must never be used to screen, diagnose or guide treatment, and its results only mean anything on images of the same kind it was trained on.
Model available online, runnable in the browser, with no diagnostic validity.