Physically-Based Rendering for Indoor Scene Understanding Using Convolutional Neural Networks
This work addresses the data scarcity problem for researchers and practitioners in computer vision, particularly for indoor scene applications like robot navigation, by providing insights into synthetic data generation, though it is incremental as it builds on existing synthetic data approaches.
The authors tackled the bottleneck of limited per-pixel ground truth data for indoor scene understanding tasks by creating a large-scale synthetic dataset with 400K physically-based rendered images from 45K 3D scenes, and they found that pretraining with this dataset improved state-of-the-art results on surface normal prediction, semantic segmentation, and object boundary detection.
Indoor scene understanding is central to applications such as robot navigation and human companion assistance. Over the last years, data-driven deep neural networks have outperformed many traditional approaches thanks to their representation learning capabilities. One of the bottlenecks in training for better representations is the amount of available per-pixel ground truth data that is required for core scene understanding tasks such as semantic segmentation, normal prediction, and object edge detection. To address this problem, a number of works proposed using synthetic data. However, a systematic study of how such synthetic data is generated is missing. In this work, we introduce a large-scale synthetic dataset with 400K physically-based rendered images from 45K realistic 3D indoor scenes. We study the effects of rendering methods and scene lighting on training for three computer vision tasks: surface normal prediction, semantic segmentation, and object boundary detection. This study provides insights into the best practices for training with synthetic data (more realistic rendering is worth it) and shows that pretraining with our new synthetic dataset can improve results beyond the current state of the art on all three tasks.