Training Data Generating Networks: Shape Reconstruction via Bi-level Optimization
This addresses shape reconstruction for computer vision applications, but it appears incremental as it builds on existing bi-level optimization and few-shot learning methods.
The paper tackles 3D shape reconstruction from a single image by proposing a novel representation where a network generates training data for another learning algorithm, modeled via bi-level optimization, and reports improvements on standard benchmarks.
We propose a novel 3d shape representation for 3d shape reconstruction from a single image. Rather than predicting a shape directly, we train a network to generate a training set which will be fed into another learning algorithm to define the shape. The nested optimization problem can be modeled by bi-level optimization. Specifically, the algorithms for bi-level optimization are also being used in meta learning approaches for few-shot learning. Our framework establishes a link between 3D shape analysis and few-shot learning. We combine training data generating networks with bi-level optimization algorithms to obtain a complete framework for which all components can be jointly trained. We improve upon recent work on standard benchmarks for 3d shape reconstruction.