Identifying semiconductor manufacturer from images with convolutional nets

I’d like to showcase simple approaches to use convolutional nets — a fundamental building block for deep learning in computer vision — to identify the manufacturer of a semiconductor chip. Semiconductors looks fairly similar when photographed, and telling them apart can be difficult from just an image. I recently implemented a simple approach to fine-tune pre-trained convnets that solves the task quite well. I’d like to briefly go over the approach I used, how I set up experimentation etc.