Classify each lens model as Successful or Failed based on the object cutouts and modeling output panel. Refer to the Criteria button at any time for the full evaluation criteria. You can press Back to revise a previous answer. Failures will be characterised in detail in Round 2.

This round covers every object you marked as Failed. For each, select which panel or panels contain the problem and provide a description of what you observe.

For each object you will see the object cutouts, the modeling output, and the segmentation maps. Refer to the Criteria button for examples of common failure modes.

These descriptions will be used to optimize both the modeling and segmentation pipelines.

This round evaluates a small subset of objects to identify which parts of the inspection panels carry the most diagnostic signal for lens model quality assessment.

When scaling this validation to thousands of objects, knowing which panels and columns are essential allows us to reduce inspection overhead without loss of information. For each object, select the individual columns you used when forming your judgement in Round 1.

Use the notes field for any additional observations on panel usefulness.