Poster: On the Feasibility of Training Neural Networks with Visibly Watermarked Dataset

TL;DR

This paper introduces DeepStamp, a generative adversarial network framework that creates visually perceptible, robust watermarked images suitable for training image classification models with minimal accuracy loss.

Contribution

DeepStamp is a novel watermarking method that enables watermarked images to be used effectively for training neural networks, addressing issues of visual information loss and watermark removal.

Findings

Watermarked images trained with DeepStamp maintain high classification accuracy.

DeepStamp produces watermarks that are human-perceptible and resistant to removal.

Watermarked datasets can be used as effective training sources for neural networks.

Abstract

As there are increasing needs of sharing data for machine learning, there is growing attention for the owners of the data to claim the ownership. Visible watermarking has been an effective way to claim the ownership of visual data, yet the visibly watermarked images are not regarded as a primary source for learning visual recognition models due to the lost visual information by in the watermark and the possibility of an attack to remove the watermarks. To make the watermarked images better suited for machine learning with less risk of removal, we propose DeepStamp, a watermarking framework that, given a watermarking image and a trained network for image classification, learns to synthesize a watermarked image that are human-perceptible, robust to removals, and able to be used as training images for classification with minimal accuracy loss. To achieve the goal, we employ the generative multi-adversarial network (GMAN). In experiments with CIFAR10, we show that the DeepStamp learn to transform a watermark to be embedded in each image and the watermarked images can be used to train networks.