Photocard Verifier

Upload a Bangla news photocard — the model judges whether it is Real or Fake.
YOLOv8-cls 5-fold CV ≈ 97.4%
📤
Click or drop an image here
or paste with Ctrl / Cmd + V
preview
🔍 Result will appear here after you analyze an image.
Confidence
📄 ⚡

📖 How to use

  1. Add a photocard. Drag an image onto the upload box, or click it to pick a file. A photocard is the square news graphic shared on Facebook: a photo with a Bangla headline laid over it. JPG, PNG, WEBP, GIF and BMP work, up to 15 MB.
  2. Press Analyze. The image is resized to 256×256 and passed through the model. It normally takes under a second; the first request after the Space wakes can take a few seconds longer while the weights load.
  3. Read the verdict. You get Real News or False News plus a confidence bar. High confidence means the image looks strongly like the examples in that class, not a probability that the story is true.
  4. Tell us if it is wrong. Use 👍 / 👎 under the result. A 👎 lets you give the label you believe is correct and an optional note. Corrections are stored and used to improve later versions.
What this model actually judges. It was trained on the visual style of Bangla news photocards (layout, typography, logo treatment, compression artefacts), so it spots cards built to imitate a real outlet. It does not read the Bangla text or fact-check the claim, so a fabricated story on a genuine outlet's template can still come back as Real News. Treat the verdict as one signal and confirm anything important against the outlet directly.