Finally starting AI training!!
After weeks of annotating images, I can finally start training my model on my dataset.
This project is a CV model that can detect fire and smoke, which I can open-source on GitHub or run on an edge device such as a Raspberry Pi. My original plan was to use RT-DETR, but due to its heavy training time (6 Hours for a single Epoch!), I decided to shift to YOLOv8 instead, which is much better suited for edge deployment and is easier on my laptop.
I collected the data from scraping the web, which took a lot of time in itself. Thankfully, this isn’t my first rodeo, as I have collected and trained AI before, but not at THIS scale (about 15k images!!)
I plan to apply it in security cameras for fire detection in open spaces and maybe even in drones for wildfire detection.
This took longer than expected due to exams, school, and coaching. It will still take hundreds of hours, but in that time, I can focus on several other projects of mine.
If you have any questions, feel free to reach out and stay tuned!!
