Open ChCh1999 opened 8 months ago
Under the graph constraint, the evaluator is supposed to take only one prediction for each subject-object pair. However, in the implement of MEET, you just concatenates the pred_rel_scores and pred_rel_labels of each group. That means the evaluator will include 5 prediction for each subject-object pair, which might result in an unfair comparison. Have you noticed this in your experiments? https://github.com/visinf/veto/blob/832ba0e4f79595fad60564a9f2409472b84a15a9/pysgg/modeling/roi_heads/relation_head/inference.py#L376-L397
pred_rel_scores
pred_rel_labels
Under the graph constraint, the evaluator is supposed to take only one prediction for each subject-object pair. However, in the implement of MEET, you just concatenates the
pred_rel_scoresandpred_rel_labelsof each group. That means the evaluator will include 5 prediction for each subject-object pair, which might result in an unfair comparison. Have you noticed this in your experiments? https://github.com/visinf/veto/blob/832ba0e4f79595fad60564a9f2409472b84a15a9/pysgg/modeling/roi_heads/relation_head/inference.py#L376-L397