BlueVetaUpright

BlueVetaUpright is an API that detects whether a product photo is sideways or upside-down and returns the exact fix: the 4-axis rotation (0/90/180/270 degrees) plus a precise fine-tilt angle, each with its own calibrated confidence score.

Built as a purpose-trained model, not a general vision model, specifically for this one task. Independently measured accuracy: 99% on clean quadrant rotations, 97% across the full realistic range including fine-tilt correction, and 84% on the hardest cases (severe tilt, no clean quadrant to anchor on). It outperforms general-purpose vision models like ChatGPT and Gemini on this specific task.

Send a photo as raw image bytes or a file upload; get back a JSON response with the rotation fields, a confidence score, and an honest flag when the answer should get a second look instead of being auto-applied.

Response fields include the 4-axis rotation and its confidence, the full rotation (4-axis plus fine-tilt) and its own confidence (honestly lower than the 4-axis confidence on purpose), a resolved boolean, and a plain-English note (low_confidence or exif_disagreement) when something needs a second look. HEIC photos from iPhones are decoded automatically before analysis. No separate test mode -- every call returns a real, usable result.

Typical use cases: catching sideways or upside-down product photos before they reach a marketplace listing, normalizing bulk user-uploaded photos, and flagging low-confidence cases for manual review instead of silently guessing.

Priced simply: 3 cents per image, pay as you go, no subscription.