Flickr API/XML

Another reason choosing Flickr made sense

One of the best things about Flickr is their open API which allows for access to image information as XML data. This allowed me to grab all of the photo information in the Squared Circle group and filter it by what Creative Commons license it was associated with. So along with the images themselves I also have all the tags and titles, which I might use somehow in the future.

xml

 

Downloading

I need 16,384 images please

Now that I knew the URLs of the images I wanted to work with I needed to get them off of Flickr and on to my local hard drive. I discovered a fabulous program called DeepVacuum which lets takes a text file full of URLs and downloads one every few seconds so it doesn't overload the server. It worked perfectly and a half-day later or so I had all the images I needed to work with.

Defining the Dimensions

Getting to know the images

I chose 10 different dimensions to describe each image:

  • Brightness
  • Saturation
  • Redness
  • Greeness
  • Blueness
  • Internal Dissimilarity (from JIU)
  • Horizontal Edginess (Sobel Filter)
  • Vertical Edginess (Sobel Filter)
  • Top Pointing Diagonalness (FFT)
  • Bottom Pointing Diagonalness (FFT)

An example of the image filters mentioned above is below. After applying the filter I then counted the brightness of each pixel to determine the value of that dimension.

filters

All image analysis was done with Processing. Simple dimensions such as brightness went quickly but more complicated ones such as internal dissimilarity took long enough to warrant an overnight run. Sobel and FFT processing was done ahead of time using Photoshop so that the images could be analyzed multiple times without recalculating the filter. The other reason Photoshop batch processing was used was because of time constraints and the trouble I was having getting it to work on the fly in Java. I want to learn how to do these calculations from within Java but quick and dirty batch processing had to come to the rescue for now.