Deep Learning #

This class was certainly challenging. It was quite homework heavy and focused mainly on computer vision models. We created classifiers, detectors, segmentation models, and GANs from scratch in our homeworks. While it was quite stressful and confusing when trying to debug problems with the model, I became quite comfortable with PyTorch and distributed training.

Since a lot of the homeworks were based on famous open source models, I became quite good at learning from machine learning papers and adapting their code and insights to my models.

Our final project was quite open-ended (anything computer vision related). My group and I decided to work on building a model that could help drones navigate terrain with low-visibility. It was a lot of fun doing a literature review, experimenting, and obtaining interesting results. You can find more infomation here

I’ll flush this out and add some pictures of grumpy cats I made using a GAN (as soon as I can find where the pictures are)