Skip to content

The issues with the evaluation script. #5

@llbll

Description

@llbll
Image
  1. When using Docker for evaluation, the WARNING shown in the above image appeared: 2025-12-24 05:48:49,034 - compute_metrics_parallel - Error evaluating constraint is_visible. Is this normal?

  2. The evaluation metrics we finally obtained have a large gap compared to those in the paper(

Image

). Is there something wrong with our configuration? The evaluation results are as follows:
{
"language_adherence_success": 15.171428571428573,
"global_constraint_accuracy": 45.08521418701059,
"complete_placement_success": 8.542857142857143,
"subgroup_metrics": {
"spatial": 0.5661352865144335,
"rotational": 0.17298735861610112,
"visibility": 0.0,
"physical": 0.5731428571428572
},
"per_constraint_metrics": {
"on": {
"accuracy": 0.9378238341968912,
"satisfied": 1267,
"total": 1351
},
"below": {
"accuracy": 0.2553191489361702,
"satisfied": 48,
"total": 188
},
"is_visible": {
"accuracy": 0.0,
"satisfied": 0,
"total": 996
},
"near": {
"accuracy": 0.49556650246305417,
"satisfied": 1006,
"total": 2030
},
"above": {
"accuracy": 0.16822429906542055,
"satisfied": 18,
"total": 107
},
"not_visible": {
"accuracy": 0.0,
"satisfied": 0,
"total": 214
},
"plausible": {
"accuracy": 0.5731428571428572,
"satisfied": 2006,
"total": 3500
},
"facing": {
"accuracy": 0.17298735861610112,
"satisfied": 260,
"total": 1503
},
"adjacent": {
"accuracy": 0.4088397790055249,
"satisfied": 74,
"total": 181
},
"between": {
"accuracy": 0.27388535031847133,
"satisfied": 215,
"total": 785
}
},
"metrics_by_constraint_count": {
"2": {
"total_placements": 1481,
"at_least_one_correct": 944,
"all_correct": 203
},
"1": {
"total_placements": 878,
"at_least_one_correct": 258,
"all_correct": 258
},
"3": {
"total_placements": 1049,
"at_least_one_correct": 848,
"all_correct": 70
},
"4": {
"total_placements": 92,
"at_least_one_correct": 89,
"all_correct": 0
}
}
}

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions