Datasets:
| { | |
| "test_set": { | |
| "n_samples": 4200, | |
| "n_non_natural": 2041, | |
| "n_natural": 2159 | |
| }, | |
| "metrics": { | |
| "roc_auc": 0.887, | |
| "average_precision": 0.8906 | |
| }, | |
| "threshold_05": { | |
| "accuracy": 0.7993, | |
| "confusion_matrix": [ | |
| [ | |
| 1548, | |
| 493 | |
| ], | |
| [ | |
| 350, | |
| 1809 | |
| ] | |
| ] | |
| }, | |
| "threshold_07": { | |
| "accuracy": 0.7526, | |
| "confusion_matrix": [ | |
| [ | |
| 1888, | |
| 153 | |
| ], | |
| [ | |
| 886, | |
| 1273 | |
| ] | |
| ] | |
| }, | |
| "training_data": { | |
| "total_labeled": 26344, | |
| "usable_after_download": 20996, | |
| "label_distribution": { | |
| "non_natural": 10323, | |
| "natural": 10673 | |
| }, | |
| "labeling_method": "active learning loop on ~200k pool from LAION-2B-en", | |
| "feature_model": "openai/clip-vit-large-patch14", | |
| "feature_dim": 768 | |
| } | |
| } |