VibrantLeaves: A principled parametric image generator for training deep restoration models
This work addresses biases in training sets for image restoration models, offering a principled synthetic alternative that enhances robustness, though it is incremental as it builds on the classical Dead Leaves model.
The paper tackled the problem of training biases in deep neural networks for image restoration by introducing a synthetic image generator based on geometric modeling, textures, and image acquisition principles. It showed that networks trained on these synthetic datasets achieve performance nearly equal to those trained on natural images, with improved robustness to geometric and radiometric perturbations.
Even though Deep Neural Networks are extremely powerful for image restoration tasks, they have several limitations. They are poorly understood and suffer from strong biases inherited from the training sets. One way to address these shortcomings is to have a better control over the training sets, in particular by using synthetic sets. In this paper, we propose a synthetic image generator relying on a few simple principles. In particular, we focus on geometric modeling, textures, and a simple modeling of image acquisition. These properties, integrated in a classical Dead Leaves model, enable the creation of efficient training sets. Standard image denoising and super-resolution networks can be trained on such datasets, reaching performance almost on par with training on natural image datasets. As a first step towards explainability, we provide a careful analysis of the considered principles, identifying which image properties are necessary to obtain good performances. Besides, such training also yields better robustness to various geometric and radiometric perturbations of the test sets.