Random Walk is a web application to allow users to identify and explore regions with specific scenery via Flickr photos. Photographs are analyzed using a deep learning model based on the AlexNet structure and trained on the MIT Places205 image set (2.5 million images).

The primary dependencies for the web server itself are Flask and Caffe. Also, if you want to take advantage of the pretrained Places model see the Caffe Model Zoo.

Download and compile pycaffe using the installation instructions, then edit the Random-Walk.py file to point to your SQL database.

Finally, run the Flask web server via

sudo python Random-Walk.py

The imclass_remote.py file is meant to work with csv files from the Flickr 100M photo set. After downloading and extracting the files the images can be classified via python imclass_remote.py PATH_TO_CLASSIFICATION_CSV PATH_TO_PHOTOSET_CSV NUMBER_OF_POINTS

Then the fileset is added to your MySQL database with python database_entry PATH_TO_CLASSIFICATION_CSV