FireDetection

A neural network that detects fire and smoke on images and videos

Scripts definition

For training or combined predictions you can run idk.py. For video prediction run autotest.py. It will predict all videos in Tests directory. For custom source prediction run test.py. You can select any links or paths for the videos.

How to run

Prerequisites

The script requires the following libraries for work (you could use pip install to install all of them):

• ultralytics
• torch
• onnx

Steps

1. Clone the repo into any specified directory (it'll be named $InstallPath/ below) using the following command: git clone https://github.com/IOExcept10n/FireDetection.git. If you encounter any problems with downloading, try Git LFS for cloning.
2. Let's test the model on any sample video (For example, this one).
3. Run test.py
4. Input the following parameters:

Precision threshold: 0.25
Experiment name: SampleTest
Model name: best_m.pt
Source path: https://www.youtube.com/watch?v=POZ4B2gPVmc

Enjoy the result (it'll be stored in path $InstallPath/runs/detect/SampleTest)

Metrics

F1

Precision

Precision-Recall

Recall

Train results

Labels