First object detection on fire forest

Hello everybody,

I am new in artificial intelligence and more precisely in computer vision. I started to train a YOLOv5 model to detect the beginnings of forest fires. I discovered Roboflow while looking for tutorials on how object detection works and I found their explanation great. Moreover the articles written are very interesting to understand some data like the mAP.

I am currently building a dataset with images of burning forest and soon I will add smoke detection. I’m thinking of making this dataset available to everyone so take advantage of it if you are interested.

For the moment I use roboflow mainly to manage my dataset with the increase of the data and also to make some retouches on my annotations which was not correct (I begin :sweat_smile:). But I think that soon I will test to train a model on my dataset directly from their web interface.

I think that regularly I will add information about this project so feel free to react to this post. I will be happy to discuss with you.


Thanks for sharing! Feel free to use these resources in case they are helpful as you work on the project :slight_smile:

Good luck and keep us posted!

1 Like

Hi @teodf - wanted to check in on this. How is the project going?

Any issues with detections, or any other questions about model improvement?

If you didn’t know, we added “Cloning” to Universe, to make it easier to add images to a dataset.

Here’s a page for “Smoke” datasets: Smoking Datasets and Pre-Trained Models

And for “Fire Detection” (subject results): fire detection | Roboflow Universe Search

And for “Fire Detection” (metadata results): fire detection by Metadata | Roboflow Universe Search

Oh - and since you said you’re new to computer vision, we have some good resources for you at Sign in to Roboflow and Learn Computer Vision with Roboflow