Back to evaluations
Public evaluation

vision_defect_eval

Evaluates accuracy of vision defect taxonomy classification from bounding box telemetry.

Evaluation type
task based
Challenge
Machine Vision Quality Inspection Agent with Mastra AI
Difficulty
Intermediate
Rigor
Unspecified

Evaluation overview

How the linked challenge is judged: tasks, benchmarks, and criteria count.

Tasks
1
Benchmarks
0
Criteria
0

Task templates

Inputs and expected outputs.

Task 1

vision_defect_eval

Classifies bounding box feature telemetry into defect categories.

Input format

JSON containing bounding box dimensions and surface reflection intensity

Output format

JSON defect category, severity grade, and rejection flag