Iteratively Trained Interactive Segmentation
This work addresses the need for more efficient data labeling in computer vision, though it appears incremental as it builds on existing interactive segmentation approaches.
The paper tackles the problem of expensive manual labeling for object segmentation by developing an interactive system that uses user clicks as input to a convolutional network, proposing an iterative training strategy that adds clicks based on segmentation errors during training, which results in improved performance over state-of-the-art methods.
Deep learning requires large amounts of training data to be effective. For the task of object segmentation, manually labeling data is very expensive, and hence interactive methods are needed. Following recent approaches, we develop an interactive object segmentation system which uses user input in the form of clicks as the input to a convolutional network. While previous methods use heuristic click sampling strategies to emulate user clicks during training, we propose a new iterative training strategy. During training, we iteratively add clicks based on the errors of the currently predicted segmentation. We show that our iterative training strategy together with additional improvements to the network architecture results in improved results over the state-of-the-art.