Learning to Estimate Indoor Lighting from 3D Objects

In this work, we propose a step towards a more accurate prediction of the environment light given a single picture of a known object. To achieve this, we develop a deep learning method that is able to encode the latent space of indoor lighting using few parameters that is trained on a database of environment maps. This latent space is then used to generate predictions of the light that are both more realistic and accurate than previous methods. Our first contribution is a deep autoencoder which is capable of learning the feature space that compactly models lighting. Our second contribution is a convolutional neural network that predicts the light from a single image of a known object. To train these networks, our third contribution is a novel dataset that contains 21,000 HDR indoor environment maps. Finally, we evaluate our method on a dataset of synthetic objects and find it to outperform the state-of-the-art techniques across a variety of materials and poses.