Mayb is context-aware image recommendation system specified on the pinterest. This system proposes related images in addition to similar images on the basis of repins of pinterest users. The point is to make use of context associated with image without analysing image itself.

Crawler is not working correctly now, because the specification of pinterest has changed. Fortunately, I seem to be able to fix it. Thanks. See you later.

There is variation in the way of users's classification of images. It is pinned to the different board even in the same image. This feature can be used to find related images using pearson's correlation.

Maby is depend on these libraries. You have to install those with easy_install.

• lxml 2.3.5
• chardet

$ sudo easy_install "lxml==2.3.5"
$ sudo easy_install chardet

In the beginning of recommending images, we have to setup database.

$ cd /path/to/src
$ python mayb.py setup

option to crawle pinterest is:

$ cd /path/to/src
$ python mayb.py crawle

option to cluster pins we got is:

$ cd /path/to/src
$ python mayb.py train

In the end, run the following command in your new terminal and visit at http://localhost:8000/. The images which is displayed on are ones that Mayb could recommend to you. If you click any image, Mayb would shows you similar images which has strong relationship with that you select.

$ cd /path/to/www;
$ python -m CGIHTTPServer

• run python mayb.py crawle before run python mayb.py train.