The PhotoBook Dataset
Segmentation Evaluation and Validation
Segment Chain Data Set

The PhotoBook Dataset

The PhotoBook Dataset can be cloned from the Github repository at

Alternatively, the files can be downloaded directly by clicking the following links:

The PhotoBook Logfiles (zip-compressed, 8MB)

The PhotoBook Image Sets (zip-compressed, 60MB)

The PhotoBook Processor (Python class, 5kB)

The ACL Analysis

The script to replicate the analysis presented here and in our ACL paper can be viewed here and downloaded from here. Running the notebook requires the log folder and the out-of-vocabulary dictionary oov_dictionary.pickle available here.

Executing the analysis script produces a number of pandas dataframes that can be downloaded as pickled files from github as well, or directly by clicking the following links:

dataset.pickle (pickled, 65.5 MB)

game_statistics.pickle (pickled, 1.12 MB)

worker_data.pickle (pickled, 514 kB)

The Segment Chain Dataset

The segment chain dataset creator can be downloaded from github here, or as a zipped archive here (zipped, 6.3MB). To execute the files, place the log folder into the project's data folder. Further documentation can be found here.

The produced output files can be found here (zipped, 5.59MB).

The Discriminator

All files necessary for running the image discriminator model as presented in the ACL paper can be found on github here. The pre-trained models that produced the numbers shown in the paper can be downloaded from github here.
The README for the discriminator models can be viewed here.