Skip to content
Rubric Help

Photo Logging (Computer Vision)

Photo Logging (Computer Vision)

Photograph your meal and walk away. The AI analyzes in the background while you eat.

Photo logging is designed around a “capture now, reconcile later” workflow. The goal is to get the photo taken in under a second so logging never interrupts your meal.

The Workflow

  1. Tap the camera button on the Log screen.
  2. Take a photo of your meal. The camera captures instantly — there is no preview step, no filters, no cropping. The design goal is shutter-to-done in under 500 milliseconds.
  3. Return to your day. The photo uploads and the AI processes it in the background, identifying food items and estimating nutrition for each one.
  4. Review later. When processing is complete, the entry appears in your Open Items queue. Open it to see what the AI detected, adjust if needed, and close the entry.

Open Items

Open Items is your queue of entries waiting for review. After the AI finishes analyzing a photo, it appears here with:

  • The original photo
  • Detected food items with individual nutrition estimates
  • A combined total for the meal
  • Correction tools (the Universal Resolver)

You can review and close entries at any time. When you close an entry, the nutrition values are finalized and the photo is deleted from Rubric’s servers.

Tips for Better Photo Results

Shoot from above. An overhead angle (straight down at the plate) gives the AI the best view of all items.

Use good lighting. Natural light works best. Harsh shadows and dim restaurant lighting reduce detection accuracy.

Include the entire plate. Make sure all food is visible in the frame. Partially hidden items may be missed.

One meal per photo. Photograph a single plate or tray. Combining multiple people’s meals in one photo produces unreliable results.

Managing Expectations

Computer vision works well for distinct, visible food items — a plate of chicken, rice, and vegetables; a sandwich with identifiable components; a salad with distinguishable toppings.

It works less well for mixed dishes where individual ingredients are not visible: casseroles, smoothies, thick soups, heavily sauced pasta. For these, text logging (“beef stew with potatoes and carrots, about 2 cups”) often produces better estimates.

When the AI gets it wrong, the Universal Resolver gives you precise correction tools. This is not a flaw in the system — it is the system working as designed. Fast capture with accurate correction tools outperforms slow, friction-heavy input.

Privacy

Photos are handled with strict privacy controls:

  • Location data and other metadata (EXIF) are stripped on your device before the photo leaves your phone.
  • Photos are deleted from Rubric’s servers after you close the entry.
  • Rubric does not build a photo archive of your meals.

For full details on how Rubric handles your data, see the Privacy article.

Related Articles