This week I spent most of my time resolving the python issue that plagued my computer, followed by a robotics competition I had to leave for on Friday. Unfortunately, I don’t remember enough of the details to give a step by step guide, but that probably doesn’t matter anyway as situations differ and my problem is probably not super replicatable. I will say that if you are installing python on your computer in the future, install it in a virtual environment so it’s easily removable. Alternatively, stick with anaconda right from the start, which is what I ended up with on my system – removing all instances of python from my computer and reinstalling just anaconda.
Since this was finished towards the tail end of my week, I didn’t end up working much on my neural network modification. However I was able to work with my advisor and we found a very good dataset from Berkeley that suited our purposes very well. This is Berkeley’s BAIR dataset (https://bair.berkeley.edu/blog/2018/05/30/bdd/) which consists of a series of images taken from video footage driven around New York, San Francisco, Berkeley, and the general Southern Bay Area. What makes this dataset so good is that it was taken over many many days and many many cities, which means the pictures that are rainy are quite varied from each other. Plus, there is a sub-dataset (which is what I will actually be using) with pulled images every few minutes from the video footage. So thankfully, I don’t have to sort through almost 2TB for my choice of rainy images. However I do have to sort through the sub-dataset for just the rainy images, along with some random non-rainy images to compare against for training. Overall not a bad week, albeit a slow one.