Gun / Knife Detection - YOLO11n
This repository contains the final YOLO11n model trained for gun and knife object detection.
Classes
Dataset
The model was trained on a custom merged gun/knife YOLO dataset.
Final dataset statistics:
| Split | Images |
|---|
| Train | 29,446 |
| Validation | 3,681 |
| Test | 3,681 |
Overall class distribution:
| Class | Images | Instances |
|---|
| gun | 23,694 | 25,333 |
| knife | 13,103 | 13,324 |
Training
- Model: YOLO11n
- Epochs: 50
- Image size: 640
- Batch size: 16
- Device: NVIDIA GeForce RTX 3050 Ti Laptop GPU
Test Results
| Class | Precision | Recall | mAP50 | mAP50-95 |
|---|
| all | 0.942 | 0.934 | 0.964 | 0.698 |
| gun | 0.970 | 0.959 | 0.984 | 0.748 |
| knife | 0.915 | 0.909 | 0.944 | 0.647 |
Intended Use
This model is intended for academic demonstration and comparative evaluation in a computer engineering graduation project.
It should not be used as a standalone safety-critical detection system without additional validation.
