I've got my inspiration from reading articles about CNNs and image processing algorithms.

What it does

My code analysis every picture using multiple GPUs for faster and more accurate results. The ResNet, or residual network, takes the residual weights from the previous layers to improve the next. Basically all is recycled and nothing is thrown away, so that he can improve the accuracy.

How I built it

I had a long session of brainstorming, and after i read all the documentation, I've decided to use the ResNet50 architecture to solve this task.

Challenges I ran into

I had a couple of errors and it took some time to implement the multi-processing Cuda based architecture.

Accomplishments that I'm proud of

It was quite a journey I've been through this few couple of days. I am really proud that i can see my code running and I've learnt so much by doing so.

What I learned

I had a huge issue with problem fixing, it was in deed a challenge, i had to do reverse engineering to correct them. Quite a carousel if I can say so, with lots of ups and downs, but in the end, we all are the winners for completing this amazing task.

What's next for ResNet50

For the moment i don't know, but maybe next year the challenge would be to improve it.

Built With

Share this project: