Object Discovery with a Copy-Pasting GAN
This addresses the problem of unsupervised object segmentation for computer vision applications, representing a novel approach rather than an incremental improvement.
The paper tackles unsupervised object discovery by proposing a copy-pasting GAN framework where the generator learns to segment objects by compositing them between images to fool a discriminator. The method achieves strong performance on four diverse datasets with challenging cluttered backgrounds.
We tackle the problem of object discovery, where objects are segmented for a given input image, and the system is trained without using any direct supervision whatsoever. A novel copy-pasting GAN framework is proposed, where the generator learns to discover an object in one image by compositing it into another image such that the discriminator cannot tell that the resulting image is fake. After carefully addressing subtle issues, such as preventing the generator from `cheating', this game results in the generator learning to select objects, as copy-pasting objects is most likely to fool the discriminator. The system is shown to work well on four very different datasets, including large object appearance variations in challenging cluttered backgrounds.