Superpixel Segmentation via Convolutional Neural Networks with Regularized Information Maximization
This is an incremental improvement for computer vision tasks like image segmentation, offering flexibility in superpixel generation without labeled data.
The paper tackles unsupervised superpixel segmentation by optimizing a randomly-initialized CNN at inference time to generate superpixels from a single image without labels, achieving advantages such as adaptive superpixel counts and controllable properties, verified on BSDS500 and SBD datasets.
We propose an unsupervised superpixel segmentation method by optimizing a randomly-initialized convolutional neural network (CNN) in inference time. Our method generates superpixels via CNN from a single image without any labels by minimizing a proposed objective function for superpixel segmentation in inference time. There are three advantages to our method compared with many of existing methods: (i) leverages an image prior of CNN for superpixel segmentation, (ii) adaptively changes the number of superpixels according to the given images, and (iii) controls the property of superpixels by adding an auxiliary cost to the objective function. We verify the advantages of our method quantitatively and qualitatively on BSDS500 and SBD datasets.