Using ML to identify packaging anomalies.

My experience

I recently partnered with the innovation lab of one of the largest biotech companies with the objective of using machine learning techniques to identify packaging anomalies on a conveyor belt. I've had to be very creative to improve the accuracy of my machine learning model. For example, I was only given a small dataset of 200 images, but I used image augmentation to expand my dataset to several thousand images by flipping, rotating, and translating images. I also had to be careful of the size of my model because it had to run in less than 100 milliseconds.

By the end of my internship, I successfully built a convolutional neural network with 99.7% validation accuracy.

Let's keep in touch.

© 2018, made by Nikhil D'Souza.