Android Food Image Nutrition Analyzer
Budget: ₹750 – ₹1,250 INR
I want to turn a smartphone camera into a pocket-sized dietitian. The job is to build an Android app that accepts a photo of a meal—whether it is a home-cooked plate, a bowl of fruit, a salad, or even a packaged snack—and returns an estimated nutritional profile. At minimum the report must list calories, carbohydrates, fats, and proteins; fibre and key micronutrients are a welcome bonus if your model supports them.
Core workflow I have in mind
1. User snaps or imports a food image.
2. The app detects and segments each food item in the photo.
3. It matches each segment to a nutrition database and scales the values by estimated portion size.
4. Results are shown clearly on-screen and saved locally for later review.
Technical direction
• Target platform: Android only.
• Feel free to leverage TensorFlow Lite, ML Kit, PyTorch Mobile or any other on-device vision stack you are comfortable with, as long as the APK remains lightweight and runs without cloud connectivity.
• A small cloud fallback for edge cases is acceptable, but the core analysis should work offline.
• Source code, model files, and a brief README explaining model training or fine-tuning steps form the final deliverables.
Acceptance criteria
• Image-to-nutrition pipeline completes in under 5 seconds on a mid-range phone.
• Mean nutrient error <20 % when tested on a held-out sample set that I will provide.
• Clean, intuitive Android UI built with modern design guidelines (Material 3 preferred).
• All third-party assets and datasets are properly credited and licensed.
If you have previous work in food recognition or nutrition tracking, mention it when you reply; it will help me gauge the suitability of your approach.
Core workflow I have in mind
1. User snaps or imports a food image.
2. The app detects and segments each food item in the photo.
3. It matches each segment to a nutrition database and scales the values by estimated portion size.
4. Results are shown clearly on-screen and saved locally for later review.
Technical direction
• Target platform: Android only.
• Feel free to leverage TensorFlow Lite, ML Kit, PyTorch Mobile or any other on-device vision stack you are comfortable with, as long as the APK remains lightweight and runs without cloud connectivity.
• A small cloud fallback for edge cases is acceptable, but the core analysis should work offline.
• Source code, model files, and a brief README explaining model training or fine-tuning steps form the final deliverables.
Acceptance criteria
• Image-to-nutrition pipeline completes in under 5 seconds on a mid-range phone.
• Mean nutrient error <20 % when tested on a held-out sample set that I will provide.
• Clean, intuitive Android UI built with modern design guidelines (Material 3 preferred).
• All third-party assets and datasets are properly credited and licensed.
If you have previous work in food recognition or nutrition tracking, mention it when you reply; it will help me gauge the suitability of your approach.