Computer vision has transformed the way we look at the world’s most challenging problems (see what I did there, look).

Despite having new algorithms and model architectures come up and change the playing field, we still have a long way to go in terms accessibility, efficiency, and accuracy of these models before we decide to have all of our doctors on the GPU.

For now though, let’s look at what can be done with some stable internet connection, a computer, and PyTorch. The result: a CNN that can accurately diagnose diabetic retinopathy.

I’m not going to focus all that much…

Learn about how you can design specific application-based neural networks in practice.

Deep learning is difficult, to say the least. The plurality of structures and methods for creating neural network architectures are almost exclusive to very selective problems. In most cases, the boilerplate models just won’t cut it, so how do you go about designing the best neural network?

Credit -> Tongyang Xu: A take on FDM Signals (in communications) where neural networks are built in similar fashions as SEFDM signals where the bandwidth is split and stems out into multiple other streams for communication.

The reality is there is no right answer, but there are a few general rules to follow founded from research in fields like sparsity, memory, pruning, to efficiently test and produce a viable model.

Looking at Neural Network Architecture (NNA) from a Different Lens

In essence, neural networks are just structures that information flows through. Certain operations are done on this information in…

I’ve always looked at Disney as this magical company with a life-like personality. It’s crazy to think that there is a corporation out there that fundamentally changed the way we kids thought about the world and the people we looked up to. As someone who has the maturity level of a peanut but also growing out of G-rated movies, I don’t often think of a company when I hear the words expanse, creativity, and culture as much as Disney.

Feature Maps of Image1 from dataset that was used in Pt. 1 Using VGG16 Model

We’re back with CNNs and developing accurate classification models. I recommend you read the last part (and clap it up too) so that you’re familiar with the content, the approach, and dataset that we’re using.

To give you a refresher, we designed a CNN to diagnose Diabetic Retinopathy from a dataset with 3662 images of patients with either Mild, Moderate, Severe, Proliferate DR, or No DR cases. After training and evaluating the model however, the accuracy only reached 23% as the loss did not go down and stayed around 1.6 [loss.item()].

After some reconstruction and learning_rate graphs, I found that…

We all wish to be different or unique in our lives, hoping to be seen for our strengths from the rest. However, some of us are born not with rare gifts, but unfortunately with rare diseases. More than 350 million people across the world are living with 1 of the 7000 rare genetic diseases (RD) discovered, 75% of which are children. Despite these staggering numbers, rare genetic disorders are severely underemphasized in the medical field as each disorder must follow an extensive and expensive detecting and treatment process that many patients across the world do not have access to.


Dev Patel

TKS Innovator | Grade 10 Machine Learning Developer | Biotech Enthusiast | Looking to learn, grow and build the ideas of the future into reality.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store