People Line Crossing Counter (YOLOv8 + Streamlit)

This project provides a complete web-based system to detect and count people crossing a custom-defined line in a video using YOLOv8 and ByteTrack. It features a Streamlit frontend and a Python backend with logging and video annotation.

---

🔧 Features

• 🖼️ Upload video and draw counting line on the first frame
• 🧠 YOLOv8 + ByteTrack to detect & track people (class_id = 0)
• 🧮 Count people who cross the defined line (with deduplication)
• 📦 Output annotated video with bounding boxes, tracker IDs, and line overlay
• 📊 Export all detection logs to Excel (bbox, class, confidence, frame, tracker_id)
• 📉 Live frame-by-frame processing progress in Streamlit

---

🛠️ Tech Stack

Layer	Technology
Object Detection	YOLOv8 (Ultralytics)
Tracking	ByteTrack
UI	Streamlit
Drawing	streamlit-drawable-canvas
Frame Handling	OpenCV
Annotation	Supervision
Logging	Excel export (via pandas)

---

🗂️ Folder Structure

project/
├── app.py                     # Main Streamlit UI
├── draw_line.py              # Drawing canvas logic
├── yolov8_track_count.py     # Core detection & counting logic
├── output/                   # All generated outputs
│   ├── <video_name>_result.mp4
│   ├── <video_name>_line_coords.json
│   └── <video_name>_detections.xlsx

---

▶️ How to Use

1. Launch the Streamlit app:

   streamlit run app.py

2. Upload a video (MP4 or AVI format)

3. Draw a line across the first frame (this is the crossing line)

4. Run detection – people crossing the line will be counted

5. Download Excel file to view all detections and bounding boxes

---

📋 Excel Output Columns

Column	Description
frame	Frame number
tracker_id	Unique person tracker ID
class_id	Object class (0 = person)
conf	Detection confidence score
x1, y1, x2, y2	Bounding box coordinates

---

⚠️ Notes

• Only class_id = 0 (person) is counted
• The same person is counted only once when crossing the line
• All logic assumes one custom-drawn line per video
• If using older versions of supervision, the direction-based counting is disabled

---

📦 Setup

pip install -r requirements.txt

Requirements (manually ensure):

• ultralytics
• opencv-python
• pandas
• supervision >= 0.6.0 (for direction-aware line zones)
• streamlit
• streamlit-drawable-canvas
• openpyxl (for Excel export)

---

📍 Next Steps

• Support multiple lines/zones
• Add filtering (e.g., count only adults, helmets, etc.)
• Heatmap visualization from detection logs

---

👤 Author

Developed by LTA Data Scientists Team