Is there a snake in this photo?

Upload a picture. The model estimates snake vs not-snake probabilities; the plain-English line is just how we apply the demo's cutoff.

How to read this

• Scores are the model's two outputs (they add up to 100%).
• Likely a snake / Likely not a snake uses a fixed cutoff from the deployment config (same rule as the project CLI when thresholds match).
• Wide or zoomed-out photos: the image is resized to a small square before scoring, so a snake that only covers a few pixels can be easy to miss. The training data skews toward clearer, closer views. Bounding-box detection is a separate future step, not this demo.