Inspiration
Knowing how your AI model is going to perform at different resolutions is important for making the best real world AI that you can and so benchmarking a GPU's AI capabilities on current AI models is important.
What it does
This program runs yolov8 for image segmentation on images of different sizes and records the fps at each resolution.
How we built it
The program is relatively simple once you have the tkinter gui and AI model setup you just have to have the two communicate and update the image resolutions.
Challenges we ran into
Doing the calculation for average fps was a little bit tricky as we had a programming error that took me a while to spot.
Accomplishments that we're proud of
We think we will use this in the future for testing gpus we use for AI applications
What we learned
We learned about our gpus and how they function with modern AI image segmentation models
What's next for ImageClassificationBenchmark
Adding in more diagnostics for the image classifcation and doing more than just image classifcation or allowing the user to select their model
Log in or sign up for Devpost to join the conversation.