Skip to content

Latest commit

 

History

History
20 lines (11 loc) · 934 Bytes

README.md

File metadata and controls

20 lines (11 loc) · 934 Bytes

Deep Steganography

This is an implemention of the Deep Steganography model presented by Baluja, S. (2017). With this model it is possible to hide secret images within cover images, making it impossible to be deciphered by the naked eye.

Model was trained using a subset of the Tiny ImageNet database in order to fasten the training process.

Example

deep_steganogrpahy

(Yeah, this implementation sure has room for improvement. In the future I may train with larger images, train on a larger subset, vary the learning rates, experiment different batch sizes)

Simplified model topology

simplified_model_topology

References

Baluja, S. (2017). Hiding Images in Plain Sight: Deep Steganography. Neural Information Processing Systems. http://www.esprockets.com/papers/nips2017.pdf

mnmoustafa, M. A. (2017). Tiny ImageNet. Kaggle. https://kaggle.com/competitions/tiny-imagenet