Code and Software

Probabilistic Image Colorization (PIC)

Summary Download

Tensorflow implementation for Probabilistic Image Colorization - generating diverse and vibrant colorization using auto-regressive generative networks - on the CIFAR and ImageNet datasets.

  • PIC repository
    • Source code written for Python 2.6+ and 3+ and Tensorflow
    • Pretrained models available for CIFAR and ImageNet
    • MIT license
Contact: amelie.royer@ist.ac.at

Similarity by Iterative Classification (SIC)

Summary Download

Implementation of SIC (Similarity by Iterative Classification), an unsupervised machine learning technique to estimate a similarity measures from repeated classification iterations on the data. See the following technical report for more details about the method, implementation and datasets used:

Contact: amelie.royer@ist.ac.at

Classifier Adaptation at Prediction Time

Summary Download

Python implementation of an online classifier adaptation scheme at prediction time. The capt_source_code.zip archive contains the implementation for running and evaluation the classifier adaptations, as well as for generating "realistic sequences" of object categories as described in the original work.

  • capt_source_code.zip
    • Source code written for Python 2.6+
    • Instructions and Dependencies in README file + Sphinx Documentation (Doc/ directory)
    • GNU GPL v2.0 license
    • Example of realistic query sequences precomputed for the ILSVRC2010 and ILSVRC2012 datasets (Data/ directory)
Contact: amelie.royer@ist.ac.at