Meta-Learning Initializations for Image Segmentation
This work addresses few-shot image segmentation for computer vision applications, but it is incremental as it extends existing meta-learning methods to a new domain.
The authors tackled the problem of few-shot image segmentation by meta-learning initializations, achieving state-of-the-art results on the FSS-1000 dataset and showing that meta-learned initializations provide value for few-shot tasks but are matched by conventional transfer learning beyond 10 labeled examples.
We extend first-order model agnostic meta-learning algorithms (including FOMAML and Reptile) to image segmentation, present a novel neural network architecture built for fast learning which we call EfficientLab, and leverage a formal definition of the test error of meta-learning algorithms to decrease error on out of distribution tasks. We show state of the art results on the FSS-1000 dataset by meta-training EfficientLab with FOMAML and using Bayesian optimization to infer the optimal test-time adaptation routine hyperparameters. We also construct a small benchmark dataset, FP-k, for the empirical study of how meta-learning systems perform in both few- and many-shot settings. On the FP-k dataset, we show that meta-learned initializations provide value for canonical few-shot image segmentation but their performance is quickly matched by conventional transfer learning with performance being equal beyond 10 labeled examples. Our code, meta-learned model, and the FP-k dataset are available at https://github.com/ml4ai/mliis .