Mobile Medicine OCR to JSON
Budget: ₹1,500 – ₹12,500 INR
I want an end-to-end feature that lets a user aim their smartphone camera at any medicine package or leaflet—or upload an existing photo—and receive a clean JSON file back with the detected details. The same workflow must feel identical on both iOS and Android.
Here’s what I need the module to do:
• Open the camera, auto-focus on the label, snap, and crop the region that contains the printed information.
• Accept an optional still image from the gallery and process it with the same logic.
• Recognise key text elements such as product name, dosage, batch number, expiry date, manufacturer, and any other legible data.
• Return the result programmatically as a JSON object so it can be stored or sent to our API.
Accuracy is more important than speed, but I still expect real-time feedback on focus and framing. You’re free to leverage mobile-friendly vision libraries (Google ML Kit, Tesseract, Vision Framework, etc.) or a custom TensorFlow-Lite model if that yields better results. Everything must run on-device; no cloud calls.
Deliverables:
1. Full source code for the iOS and Android implementation (native, Flutter, or React Native—use what lets you hit the quality bar fastest).
2. A short read-me that explains build steps, dependencies, and the JSON schema.
3. Sample JSON output from at least three different medicine packages to demonstrate consistency.
The job is complete when I can install the apps, scan or upload an image, and see an accurate JSON file appear in the console or a preview screen.
Here’s what I need the module to do:
• Open the camera, auto-focus on the label, snap, and crop the region that contains the printed information.
• Accept an optional still image from the gallery and process it with the same logic.
• Recognise key text elements such as product name, dosage, batch number, expiry date, manufacturer, and any other legible data.
• Return the result programmatically as a JSON object so it can be stored or sent to our API.
Accuracy is more important than speed, but I still expect real-time feedback on focus and framing. You’re free to leverage mobile-friendly vision libraries (Google ML Kit, Tesseract, Vision Framework, etc.) or a custom TensorFlow-Lite model if that yields better results. Everything must run on-device; no cloud calls.
Deliverables:
1. Full source code for the iOS and Android implementation (native, Flutter, or React Native—use what lets you hit the quality bar fastest).
2. A short read-me that explains build steps, dependencies, and the JSON schema.
3. Sample JSON output from at least three different medicine packages to demonstrate consistency.
The job is complete when I can install the apps, scan or upload an image, and see an accurate JSON file appear in the console or a preview screen.