Internal

Accuracy dashboard

Tests

0

0 food measurements

Mean abs % error

–

Median abs % error

–

Mean abs error

–

Within ±5%

–

Within ±10%

–

Within ±20%

–

Within ±30%

–

Key questions

Does a second photograph improve accuracy?

Single photono data
Two photosno data

Does a reference object improve accuracy?

No referenceno data
With referenceno data

Top-down vs 45-degree?

45°no data
Top-downno data

Which foods are hardest?

No data yet.

Percentage error distribution

No data yet.

Confidence calibration

ConfidenceTestsAvg error±20%In range
High0–––
Medium0–––
Low0–––

Useful confidence means High has clearly lower error than Low.

Prediction range calibration

–

Range coverage — actual weight inside the AI's predicted range (target ≈ 80%).

Too narrow
0
Reasonable
0
Too wide
0

Too narrow = actual fell outside the range. Too wide = contained the truth but spanned more than 80% of the estimate.

By food

No data yet.

By number of photos

No data yet.

By reference object

No data yet.

By photo angle

No data yet.

By lighting

No data yet.

By plate size

No data yet.

By device

No data yet.

By model

No data yet.

Structured experiments

None yet. Start one from the home screen.

Recent experiments

No experiments yet.