jashwanthpeddisetty0712/idol-tracker-yolo
031
Idol Tracker YOLO (Ganesh idol detection)
Object detection + tracking model fine-tuned to detect Ganesh idols (single class) for entry-only line-crossing counting at Ganesh-Visarjan event gates. Kept as a native Ultralytics checkpoint (.pt), not exported to ONNX.
Classes
Model Details
- Architecture: YOLO (Ultralytics)
- Format: PyTorch (
.pt) - Recommended confidence: 0.25
- Usage pattern: entry-only line-crossing counter — see
backend/app/services/idol_tracker/counter.pyin the visual-ai repo. Idols are static/slow-moving props, not people, so the tracker uses the full bounding-box centroid (not a head-crop) as the tracked point.
Usage
from ultralytics import YOLO
model = YOLO("idol_tracker.pt")
results = model.track("image.jpg", persist=True, classes=[0], conf=0.25)