Stop-sign (YOLOv12 + ByteTrack + HSV)

This project detects red-light violations in traffic videos using:

• YOLOv12 for vehicle detection
• ByteTrack for multi-object tracking
• HSV color analysis for traffic-light state (per ROI)

A violation is counted when the center of a tracked vehicle box crosses the drawn line while ANY traffic-light ROI is red. Violating vehicles stay highlighted in red with the label “VIOLATION” until the end of the clip. The output video is saved and then auto-opened.

---

1. Features

• Single violation line (red) — draw it once per video.
• Multiple traffic-light ROIs — add as many as you need; each is evaluated independently.
• Robust HSV color detection (downscaling, CLAHE on V channel, morphological cleanup, brightness-weighted voting).
• ByteTrack-based ID tracking to ensure center-line crossing is measured accurately.
• Final video automatically opens after processing.

---

2. Controls (Setup Screen)

• L → start drawing the line: click two points (start & end).
• R → add a traffic-light ROI: click two points (top-left & bottom-right). Repeat R to add more ROIs.
• U → undo the last ROI (or the line if no ROI exists).
• C → clear all ROIs.
• SPACE → start processing when at least one line and one ROI are defined.
• Q / Esc → quit.

---

3. Project Structure

project/
├─ main.py                  #  script
├─ requirements.txt         # dependencies
├─ bytetrack.yaml           # tracker config
├─ yolo12l.pt               # YOLOv12 

In the code, set VIDEO_IN to your video path (e.g., input/my_video.mp4). Output will be saved to output_violations.mp4.

---

Clone the Repository

git clone https://github.com/yanal154/Red_Light_Violation_Detection
cd Red_Light_Violation_Detection

4. Download YOLOv12 Model

Before running the project, you need to download the YOLOv12 weights file (yolo12l.pt) from the following link:

Download YOLOv12l.pt

Place yolo12l.pt in the project root (next to main.py).

---

5. Create & Activate a Virtual Environment

Windows (PowerShell)

python -m venv .venv
.venv\Scripts\Activate.ps1

macOS / Linux

python3 -m venv .venv
source .venv/bin/activate

To deactivate later: deactivate

---

6. Install Dependencies from requirements.txt

pip install -r requirements.txt

---

7. Run

Update VIDEO_IN in the script to point to your input video (e.g., "input/my_clip.mp4"), then:

python main.py

1. A setup window appears — draw the line (press L, click two points), then add one or more ROIs (press R, two points each).
2. Press SPACE to start processing.
3. When processing finishes, the output video (output_violations.mp4) is saved and auto-opened.

---

8. How It Works (Brief)

• Traffic-light color is inferred per ROI in HSV space with:

  • Downscaling large crops for speed/stability
  • CLAHE on the V channel (contrast-limited histogram equalization)
  • Morphological open/close to remove noise
  • Brightness-weighted pixel counts for red/yellow/green masks

• Global light state = RED if any ROI is red; else yellow/green/unknown.

• Violation rule: if the segment from previous center to current center intersects the violation line while any ROI is red, the track ID is marked as violating (box turns red and stays red).

---