Ask HN: Machine learning options for binary image classification

I have collected a set of several thousand photos from social media, all taken in the same area, some of which contain an image of a particular building. I'd like to use a classifier to determine whether or not an image contains the building - a simple yes/no label.

This seems like a job for machine learning, and I'm happy to manually train a model with up to a few hundred items, but I don't know where I should start. I have some experience with text classification (e.g. TF-IDF, Levenshtein distance) but nothing in computer vision.

I'm happy to explore paid services for this, or to dive into something like scikit-learn if it's appropriate. I would like to avoid Mechanical Turk if possible because I want to push my knowledge as well as solve a problem.