I recently involved in a project where we performed some computer vision object recognition and classification using neural networks and SVMs. It was necessary to use photos taken from an Android phone and it took a little while to get it working. Below is the solution I used, which involved creating a tiny Android application that took a photo, Base64 encoded it, then sent the data over HTTP POST to a Python BaseHTTPRequestHandler subclass which decoded the file and wrote it to disk.

Continue reading →