| """Metric calculations for multi-object tracking outputs.""" |
|
|
| from __future__ import annotations |
|
|
| from dataclasses import dataclass, field |
| from pathlib import Path |
|
|
| import pandas as pd |
|
|
| from src.tracking.tracker import TrackedObject |
|
|
|
|
| @dataclass |
| class TrackingMetrics: |
| """Collect frame-wise tracks and export summary analytics.""" |
|
|
| frame_records: list[dict[str, int]] = field(default_factory=list) |
| track_records: list[dict[str, float | int | str]] = field(default_factory=list) |
|
|
| def update(self, frame_index: int, tracks: list[TrackedObject]) -> None: |
| self.frame_records.append({"frame": frame_index, "active_tracks": len(tracks)}) |
| for track in tracks: |
| x1, y1, x2, y2 = track.bbox |
| cx, cy = track.centroid |
| self.track_records.append( |
| { |
| "frame": frame_index, |
| "id": track.id, |
| "class": track.class_name, |
| "confidence": track.confidence, |
| "x1": x1, |
| "y1": y1, |
| "x2": x2, |
| "y2": y2, |
| "centroid_x": cx, |
| "centroid_y": cy, |
| } |
| ) |
|
|
| @property |
| def frame_df(self) -> pd.DataFrame: |
| return pd.DataFrame(self.frame_records) |
|
|
| @property |
| def track_df(self) -> pd.DataFrame: |
| return pd.DataFrame(self.track_records) |
|
|
| def summarize(self) -> dict[str, object]: |
| frame_df = self.frame_df |
| track_df = self.track_df |
|
|
| if track_df.empty: |
| average_active = 0.0 |
| if not frame_df.empty: |
| average_active = round(float(frame_df["active_tracks"].mean()), 4) |
| return { |
| "total_unique_objects": 0, |
| "average_active_tracks_per_frame": average_active, |
| "track_duration_statistics": {}, |
| "frame_wise_object_counts": frame_df.to_dict(orient="records"), |
| "detection_confidence_distribution": {}, |
| } |
|
|
| durations = track_df.groupby("id")["frame"].nunique() |
| confidence = track_df["confidence"] |
|
|
| return { |
| "total_unique_objects": int(track_df["id"].nunique()), |
| "average_active_tracks_per_frame": round(float(frame_df["active_tracks"].mean()), 4), |
| "track_duration_statistics": { |
| "min_frames": int(durations.min()), |
| "max_frames": int(durations.max()), |
| "mean_frames": round(float(durations.mean()), 4), |
| "median_frames": round(float(durations.median()), 4), |
| }, |
| "frame_wise_object_counts": frame_df.to_dict(orient="records"), |
| "detection_confidence_distribution": { |
| "min": round(float(confidence.min()), 4), |
| "max": round(float(confidence.max()), 4), |
| "mean": round(float(confidence.mean()), 4), |
| "median": round(float(confidence.median()), 4), |
| }, |
| } |
|
|
| def save_tables(self, frame_counts_csv: str | Path, track_history_csv: str | Path) -> None: |
| frame_counts_csv = Path(frame_counts_csv) |
| track_history_csv = Path(track_history_csv) |
| frame_counts_csv.parent.mkdir(parents=True, exist_ok=True) |
| track_history_csv.parent.mkdir(parents=True, exist_ok=True) |
| self.frame_df.to_csv(frame_counts_csv, index=False) |
| self.track_df.to_csv(track_history_csv, index=False) |
|
|
