01 / UPLOAD
Drop in any image
JPG, PNG or WEBP up to 5MB. Metadata is stripped the moment it arrives.
scamdar-agent · step 01
[00:00]Receiving image…
[00:00]Stripping metadata
[00:01]Input normalized · 1024×1024
SCAMDAR-VISION v2.1 · ONLINE
Upload an image.
SCAMDAR SI decides.
CA · coming soon
The SCAMDAR Super Intelligence vision model reads every pixel, weighs what it finds and returns a single verdict based on Physiognomy in seconds.
Scammer· Probability of scamming you 7–10/10Maybe Scammer· Probability of scamming you 3–6/10Safe· Probability of scamming you 1–2/10
Pixel scan·Retina scan·Pattern match·Neural pass·Feature extraction·Confidence check·Label output·Pixel scan·Retina scan·Pattern match·Neural pass·Feature extraction·Confidence check·Label output·
Edge detection·Analyzing physiognomy·Cluster mapping·Signal weighting·Classification pass·Cross-check·Verdict·Edge detection·Analyzing physiognomy·Cluster mapping·Signal weighting·Classification pass·Cross-check·Verdict·
SCAMDAR-VISION / TRAINING CORPUS
Trained on 128,000+ facial scans. A 2.1B-parameter vision transformer fine-tuned on classical Physiognomy, mapping 468 facial landmarks per face.
- FACIAL SCANS
- 128,400
- LANDMARKS / FACE
- 468
- PARAMETERS
- 2.1B
- GPU HOURS
- 41,000
00 / HOW IT WORKS
Three steps. One verdict.
No accounts, no setup. Hand SCAMDAR SI an image and watch the agent work through it live.
02 / ANALYZE
SCAMDAR SI runs its vision pass
Features are extracted, pixel clusters scanned and the classification pass cross-checked.
scamdar-agent · step 02
[00:02]Extracting features…
[00:03]Scanning pixel clusters…
[00:04]Running classification pass…
03 / RESULT
Get a verdict
One of three labels, and a link you can share.
scamdar-agent · step 03
[00:05]Classification complete
SafeSCAMDAR-VISION v2.1
04 / LIVE
Fresh from the model.
The latest verdicts, updated as SCAMDAR SI finishes them. Every analyzed image lands in the public gallery.
LIVE FEEDView gallery →
Waiting for the first verdict…
- IMAGES ANALYZED
- 0
- SCAMMER DETECTED
- 0
- MODEL
- SCAMDAR-VISION v2.1
05 / FAQ
Questions, answered.
What is SCAMDAR SI?+
SCAMDAR SI is a vision model. You give it an image, it runs a classification pass and returns one of three verdicts.
How do the labels work?+
Every image gets exactly one verdict. Scammer means the model is confident the image shows a scammer. Maybe Scammer means it found partial signals. Safe means nothing was flagged.
What does the rating mean?+
It's the probability that the person in the image will try to scam you. It's scored from 1 to 10: Scammer covers 7–10, Maybe Scammer 3–6 and Safe 1–2.
How long does analysis take?+
Most images finish in a few seconds. When the model is under heavy load it can take longer. Keep your result link and the verdict will appear there automatically as soon as it is ready.
Are my uploads public?+
Only analyzed images appear in the public gallery and live feed, together with their verdict. Images that can't be processed are never shown publicly. Metadata such as location and camera details is stripped from every upload.
What files can I upload?+
JPG, PNG and WEBP images up to 5MB. To keep the model responsive, each visitor can submit up to 5 images per hour.