---
license: mit
tags:
- computer-vision
- object-detection
- yolov8
- security
- real-time
---

# Sentinel AI Crime Model (YOLOv8 Medium)

This is a custom-trained **YOLOv8 Medium** model explicitly designed to detect real-time threats from surveillance cameras.

### Model Description
The Sentinel AI model was trained on thousands of physical crime scene videos and acts as the vision engine for the **Sentinel AI Pipeline**. It is optimized to track background pedestrians while simultaneously isolating high-threat events like physical violence.

- **Developer:** Ayush Yele
- **Framework:** PyTorch & Ultralytics YOLOv8
- **Architecture:** YOLOv8 (Medium)
- **Epochs Trained:** 100

### Classes
This model predicts 4 specific macro-classes for emergency dispatch scenarios:
- `0`: `fight` (Physical altercations, assault)
- `1`: `weapon` (Knives, handguns, blunt objects)
- `2`: `violence` (Robbery, vandalism, rioting)
- `3`: `normal` (Pedestrians, standing objects)

### How to Use
You can plug this model directly into standard Ultralytics YOLO inference code:

```python
from ultralytics import YOLO

# Load the custom trained model
model = YOLO("AyushYele/Sentinel_Ai")

# Run inference on an image
results = model("surveillance_feed.jpg")
results[0].show()