I wanted to create a system where I want the computer to recognise pedestrians in real-time and implement this in an open-world RPG game Red Dead Redemption 2.
Firstly, I used a Python library for creating Computer Vision System to read the screen from the game but I was not getting the desired FPS to proceed further.
After that, I created a Python function for creating the Computer Vision System for better FPS. I was getting FPS around 10-15 from that Python library and with my function, I was able to get around 40-50 FPS.
Then trained a model to detect the main character first but was not getting the expected result additionally you need great computational power for training a model.
So I used a pre-trained model in my code and the detection accuracy was slightly better than the last.
The problem is that the pre-trained model was trained to recognise ‘real pedestrians’ whereas I need a model that is trained to recognise the in-game pedestrians.
The results of the Object Detection System:
https://www.youtube.com/watch?v=SSwZ21JpJU4
https://www.youtube.com/watch?v=3jPKGn1vUAQ
I wanted to create an automation for my smartphone where the phone will do simple tasks on its own such as disabling WiFi/Location/Bluetooth and other utilities when the phone is not in use.
Furthermore, an interactive system that is faster for providing me with information such as current weather, storage diagnostics, battery usage and others, everything right on my home screen without me having to open any apps.