MY BRITA
For the past few weeks, I’ve been deploying pre-trained models so, for this time, I finally made my first retrained model for detecting my BRITA(I honestly just didn’t have anything else to detect).
Since it’s my first time, I chose YOLOV5s which has a lot of tutorials. (Thank you, everyone)
PROCEDURES
I followed a tutorial on youtube video by Nicholas Renotte, so if you want to try it out yourself, I recommend you go check that out.
Video:
Deep Drowsiness Detection using YOLO, Pytorch and Python
EXPERIMENTING
This is the live camera I’ve recorded to test my model trained with 25 images of brita.
You can CLEARLY see how satisfied I am with the results.
VERDICTS AND THOUGHTS
To be honest, all I did was follow each step the video told me, but I was able to learn SO MUCH from it.
Before the tutorial, I didn’t exactly know how to prepare data for object detections and the procedures you have to take to implement retraining. I’m pretty sure that the procedures for other frameworks won’t be the exact same, but now I know VAGUELY how to find my way.
My next goal is to be able to run an inference with retrained models on Jetson. I was able to successfully deploy a non-retrained yolov5 model, so I went through the same procedures but I couldn’t convert my retrained model to a TensorRT engine this week.
Hopefully, I can do it in the coming weeks!
