Gun Detection

This repository contains the implementation of a gun detection algorithm using YOLOv7. The algorithm is trained on a custom dataset consisting of images and videos of people carrying guns. The model is designed to detect the presence of a gun in an image or video frame and output the coordinates of the bounding box around the detected object.

Requirements

• Google Colab account
• GPU accelerator
• Access to Google Drive

Installation

1. Clone the repository to you drive using the command

%cd /content/gdrive/MyDrive
!git clone https://github.com/WongKinYiu/yolov7

2. Download the pre-trained weights for YOLOv7 from the official repository:

!wget https://github.com/WongKinYiu/yolov7/releases/download/v0.1/yolov7x.pt
!wget https://github.com/WongKinYiu/yolov7/releases/download/v0.1/yolov7.pt

3.Install the required packages by running:

!pip install -r requirements.txt

Usage

Training

To use the gun detection algorithm, follow the steps below:

1. Create a folder named data in your working directory.

2. Inside data, create a folder named custom_data and put your images and labels in it.

3. Create a YAML file named custom_data.yaml inside data with the following format:

train: path/to/train/images
val: path/to/validation/images

nc: 1
names: ['helmet']

4. Train the model using the following command:

!python train.py --device 0 --batch-size 16 --epochs 100 --img 640 640 --data data/custom_data.yaml --hyp data/hyp.scratch.custom.yaml --cfg cfg/training/yolov7-custom.yaml --weight yolov7.pt --name yolov7-custom

5. The detect function returns the bounding box coordinates of the detected object. You can then use these coordinates to draw a bounding box around the object:

for box in boxes:
    x, y, w, h = box
    cv2.rectangle(image, (x, y), (x+w, y+h), (0, 0, 255), 2)

You can adjust the batch-size, epochs, and other parameters as needed.

Detection

1. Put the images you want to detect in a folder named testSamples inside data.

2. Run the following command to detect helmets in the images:

!python detect.py --weight /content/gdrive/MyDrive/VisionIsAllYouNeed/yolov7/runs/train/yolov7-custom/weights/best.pt --conf 0.4 --img-size 640 --source data/testSamples/

Dataset

The dataset used to train the gun detection algorithm is not publicly available due to legal and ethical considerations.

Acknowledgements

This implementation of the gun detection algorithm using YOLOv7 is based on the work of WongKinYiu, available at https://github.com/WongKinYiu/yolov7. The pre-trained weights used in this implementation are also from the same repository.

License

This project is licensed under the MIT License.