yscv-eval 0.1.8

Evaluation metrics (mAP, MOTA, HOTA) and dataset adapters
Documentation

yscv-eval

Evaluation metrics for classification, detection, and tracking. Dataset adapters for COCO, Pascal VOC, and CSV.

use yscv_eval::*;

let ap = average_precision(&predictions, &ground_truths, 0.5);
let report = classification_report(&predicted_labels, &true_labels);
println!("{}", report);

Metrics

37 public metric/eval functions across the crate:

Task Metrics
Classification accuracy, precision, recall, F1, confusion matrix, Cohen's kappa, ROC curve + AUC, average_precision, top_k_accuracy, classification_report
Detection mAP, COCO mAP, AP@IoU, precision-recall, evaluate_detections{,_coco,_from_dataset}
Tracking MOTA, MOTP, HOTA, IDF1, evaluate_tracking{,_from_dataset}
Regression MAE, RMSE, MAPE, R²
Image quality / segmentation PSNR, SSIM, dice_score, mean_iou, per_class_iou
Counting / pipeline / camera counting metrics, pipeline benchmark thresholds, camera diagnostics validation

Dataset Adapters (8)

Under src/dataset/:

  • COCO — parse_detection_dataset_coco, load_detection_dataset_coco_files
  • JSONL — detection + tracking
  • MOT — parse_tracking_dataset_mot, load_tracking_dataset_mot_txt_files
  • OpenImages — CSV pair (parse_detection_dataset_openimages_csv)
  • VOC — XML directories
  • YOLO — label directories
  • KITTI — label directories
  • WIDERFACE — TXT pair

Tests

95 tests covering metric correctness, edge cases, dataset parsing.