EENDO/SCOPE AI

INTERACTIVE RESEARCH WORKSPACE

Your image. A closer look.

Load an endoscopic image, run the model, and explore the predicted regions.

Images processed locally
01Choose an example

Start with a credited sample or select an image from your device.

02Inspect the prediction

Run analysis and compare the green candidate regions with the original.

03Explore and export

Change the threshold, observe the mask, and save the image or measurements.

ORIGINAL IMAGE INPUT
⌑Your image appears here
CANDIDATE REGIONS PREDICTION
⊙Run analysis to reveal the model’s prediction
No prediction yet—
IDCLASSMEAN MODEL SCOREBOX [x1, y1, x2, y2)AREA / PX
Candidate regions will appear after analysis.

The green mask shows selected pixels. Each box encloses a connected region of at least 16 pixels.

For research and demonstration only. Model scores are not disease probabilities. A missing candidate does not establish a normal examination. This model does not assess inflammation or ulcers.
About this browser version & image sources

This app runs the same trained Compact U-Net weights as the desktop prototype, converted to ONNX. It uses aspect-preserving resizing to 160 × 160 pixels and a single-threaded WebAssembly engine. Image decoding and numerical differences can produce small changes from the desktop app. Images above 4 megapixels are resized locally before processing; exported coordinates refer to that working image.

Sample photographs come from Wikimedia Commons. View image credits and licenses · Model details · Watch the captioned demo