Select a photo under "Image Selection" and estimate its geolocation by clicking on the world map.

Under "Image Selection" you can either choose an image from a predefined benchmark test set or upload your own photo. Your performance compared to our deep learning approach and the scene classification (probabilities for indoor, nature, urban) results are shown in the sections "Statistics" and "Scene Classification", respectively.

After you have clicked on "Guess Location" the results are calculated and presented. The world map is overlaid with a heat map, where the regions colored in red represent the most likely locations of the photo at hand. In addition, the photo parts that contributed at most to the decision are visualised (blue = “less important” to red = “very important”).

Image Selection

Photo License:
Result

Distance to ground truth or EXIF location:

Scene Classification
  • Probability indoor: ---
  • Probability nature: ---
  • Probability urban: ---
  • Predicted scene: ---
Statistics
  • Annotated images: 0
  • Rate of sucess: ---
  • Your mean error: ---
  • Model's mean error: ---
Reference

For details on the deep learning approach please check our publication:

This work was partly developed in the project "VIVA: Visual Information Search in Video Archives", which is funded by the German Research Foundation (DFG) (EW 134/4-1).