You are very welcome to submit your results to the contest!

The training set contains 180 color image tiles of size 5000×5000, covering a surface of 1500 m × 1500 m each (at a 30 cm resolution). There are 36 tiles for each of the following regions:

  • Austin
  • Chicago
  • Kitsap County
  • Western Tyrol
  • Vienna

The format is GeoTIFF (TIFF with georeferencing, but the images can be used as any other TIFF). Files are named by a prefix associated to the region (e.g., austin- or vienna-) followed by the tile number (1-36). The reference data is in a different folder and the file names correspond exactly to those of the color images. In the case of the reference data, the tiles are single-channel images with values 255 for the building class and 0 for the not building class.

For validation, we suggest to remove the first five images of every location (e.g., austin{1-5}.tif, chicago{1-5}.tif) from the training set.

The test set contains the same amount of tiles as the training set (but the reference data is not disclosed). There are 36 tiles for each of the following regions:

  • Bellingham, WA
  • Bloomington, IN
  • Innsbruck
  • San Francisco
  • Eastern Tyrol

The performance measures are:

  • Intersection over union (IoU) of the positive (building) class, i.e., the number of pixels labeled as building in both the prediction and the reference, divided by the number of pixels labeled as pixel in the prediction or the reference.
  • Accuracy, i.e., the percentage of correctly classified pixels.

These are computed for each of the regions individually (e.g., Innsbruck, San Francisco) and also for the overall test set.

To submit your results you should predict the classification maps for each of the tiles in the test set. Use the exact same file names as the input color images, and output 0/255 8-bit single-channel TIFF files (it should look similar to the reference data used for training). Then just create a zip file with the classification maps (bellingham1.tif, bellingham2.tif,…, sfo1.tif, sfo2.tif,…).

Since your output may be too large to send it to us, you can use this script to compress the images before creating you submission zip file.

When you are ready, you can submit your contribution here.

Thank you for you interest!

Comments are closed.