Assignment 10

Tagging the pictures of a wildlife photographer


Draft 3 - Tagging Pictures

Once a wildlife photographer is done importing (and maybe editing too) his photographs from the day's shoot, he would like to tag them with meta information to make searching for photos more efficient. The idea behind this design was to exploit the hierarchical nature of tags in the domain of wildlife photography (and photography in general maybe). Here is some relevant information, particular to my expert (Kalyan Varma), that helped me design this interface -

  • He always has a Geo-Tagger attached to his cameras on field trips. This device tags the geographic location (lattitude and longitude) of the camera into every photo, so every photo contains location information about where it was taken.
  • He almost always shoots in "Drive Mode", where the camera takes continous shots (upto 9 frames per second) when the shutter button is pressed. This means that he has many photos of the same subject that differ only by a little (For example - He becomes trigger happy and goes on a shooting frenzy when he spots something so as not to miss a moment).
  • His most common tags are location, kind of animal, animal and state of the animal (Eg - Karnataka, Bandipur Wildlife Sanctuary, Carnivore, Tiger, Feeding)

A first round of Geo-Tagging
Since the photos contain location information, the application sorts and lays the photos out into clusters onto a map. Each cluster / blob is clearly yet subtly distinguished from the other clusters. You can also see the actual location of where each set of photos were shot by following the rubber-band from the blob to the dot on the map. The application has a configurable threshold that adjusts the radius (in miles) that determines this clustering. The numbers associated with the locations on the map are to aid in the chronological order in which the photographer visited the areas. The big dots with more than one number (7,8) mean that the blobs 7 and 8 represent different locations were merged at some point by the user while tagging the photographs (discussed later)

Piling up similar photos
As you can see, the photos that were shot in "Drive Mode" are piled up together and represented as one photograph throughout the entire workflow. The number in the corner tells the user how many photographs are actually present in this pile. Each pile occupies the same screen space, hence the more the photos in the pile, the more cramped they are

Tagging a blob / cluster
The user can move the photos around and the blobs will cluster (merge and split) accordingly. The interface can be imagined as a fluid environment and blobs as bubbles that merge and burst as you move then around. As the application clusters the photos, each blob gets it's shade of green/brown so they can be clearly identified. Once the user has identified a particular blob that needs to be tagged, they select it, by clicking anywhere inside the blob (and turns yellow). They can then use the tagging interface in the bottom right corner (explained later). The tags themselves actually look and feel like tags (that used to hang from your new Shirt). Tags are actually attached to each blob with a loose rubber-band/string on the screen as you can see on the right. Tags can be deleted by clicking on the red icon.

Tag Hierarchy and the tagging interface
Assuming that there are tags that have been previously used in this application, and given the blob-like clustered hierarchical design, the application knows the tag heirarchy for any given tag. So if the user wants to tag the Panther cubs in the lower left blob, they simply separate out the panther cub photos from the panther blob, thus creating a new blob. They then select this blob. Once selected, the tagging interface shows the user the tag heirarchy for this blob. It shows the user what the parent tags for this photograph are - Bandipur, Karnataka, Panther, Carnivore. It also recommends to the user some tage that usually follow this heirarchy. So if you imagine all the tags to be a graph, then the application will recommend nodes in the graph that have not yet been connected in the current set of tagged photos. So the tag 'Cub' may have been recommended already. If it has not, the user simply enters the word 'Cub' and adds the tag. Assuming that 'Cub' has been entered many times before, the application can recommend the next tag / sub-tag that this blob could be tagged with - Playing, Feeding, Suckling and Grooming. The size of the recommended tags signify the confidence in recommendation / probability of the next tag / popularity of the tag. Either way, it helps the user tag entire sets of photographs without having to enter a single word. As you can see, the tagging interface uses subtle earthy colors (to match the domain of wildlife too) and the strings / rubberbands to show the relationship between the tags. The green button can be clicked to add the tag to the selected blob of photos