MIT/Tuebingen Saliency Benchmark


To submit your model to the MIT/Tuebingen Saliency Benchmark, follow these steps:
  1. Download the 300 images of the MIT300 test set.
  2. Run your model to compute predictions for each model
  3. Save the predictions to .mat, .npy, .jpg or .png files of the same size and name as the original image. If your model is probabilistic, save the log density predictions. Please save them as a zip or tar folder with the name of your model as the folder name. Alternatively you can use pysaliency's "export_to_hdf5" function.
  4. Submit your maps to Please mention whether your model is a probabilistic model predicting fixation densities, or a classic saliency map model.
  5. We post your score and model details on this page
  6. Let us know if you have a publication, website, or publicly available code for your model that we can link to your score in results table.