字幕列表 影片播放 列印英文字幕 Imagine: make a trip and take some pictures. Upload them on the web, and get this: At the bottom there are the stages of your trip: let's begin with Santiago of Chile. At the center you can see your pictures that are possible to scroll with the forward and back buttons. Scrolling the picture, at the right, your trip will be reconstructed on a map: actually, the position will be indicated for every picture that has been taken. On the left side, a tagcloud will be created in relation to the pictures that you see. Continuing with your trip, you can scroll the stage always keeping visible your way and the tagcloud associated to the picture. You can also skip some stages by choosing oneyou are interested in. Let's go to San Pedro. Unfortunately, your camera broke and you bought a new one provided with an integrated compass: now the map is able to indicate, besides the position of the picture, the direction towards the picture was taken. The first picture was taken towards north; the second one was taken towards east, and so on and so forth. The system should be able to organize your pictures without your involvement. However, you can edit the map-position and customize the tagcloud. Let's begin with the position of the first picture, taken in Santiago of Chile. After having pressed the edit button, it will be possible to edit the position of the marker doing drag&drop. Alternatively, clicking on the marker, it will be possible to associate a new address or new geographic coordinates. The same process will be possible for tags: if you are not satisfied with automatic tags associated to the picture, you can remove some of them. You can increase the ranking of the tag. You can decrease the ranking of another one. Finally, you can add your own tags! Every change you make will contribute to improve the automatic identification of the best tags for every picture. The system is smart, and able to learn! This is likely a better proposal of design. Both of them are representations of the general idea of this project. However, the interface will be definitely designed through more accurate studies. Every camera recently produced is able to date and geotag every picture. Every picture is inserted in a space-temporal context. This information is saved through adequate metadata in every jpeg picture. Uploading a set of pictures, the system is able to cover the space-temporal context of your trip. ...or of your life! During your trip, you take many pictures. During your trip, you also move around a place, continuing to take pictures. The algorithm is able to cluster these pictures considering two criteria: the geographic proximity and the temporal proximity. Every identified group is a stage of your trip. Every customization of tag-clouds associated to every picture contributes to teach the system what the most used tags are for every geographic position. For instance, if people take some pictures in the proximities of the Colosseum, in Rome, the most common word to tag these pictures will be "Colosseum". The system, in this way, is able to learn that in the proximities of these geographic coordinates, the tag "Colosseum" is very used, becoming statistically relevant. In this way every time that there is a picture in the proximities of these geographic coordinates, the system will assign automatically the tag "Colosseum" ...and those that are statistically relevant!