Deep Supervision with Shape Concepts for Occlusion-Aware 3D Object Parsing
This work addresses occlusion reasoning and holistic scene interpretation for applications in computer vision, but it is incremental as it builds on existing deep learning methods with a novel supervision scheme.
The paper tackles monocular 3D object parsing by introducing a deep CNN that localizes semantic parts in 2D and 3D while inferring visibility, using deep supervision with shape concepts to regularize the network. It achieves state-of-the-art performance on real image benchmarks like KITTI, PASCAL VOC, PASCAL3D+, and IKEA for tasks such as 2D/3D keypoint localization and instance segmentation, with effective knowledge transfer from synthetic data reducing overfitting.
Monocular 3D object parsing is highly desirable in various scenarios including occlusion reasoning and holistic scene interpretation. We present a deep convolutional neural network (CNN) architecture to localize semantic parts in 2D image and 3D space while inferring their visibility states, given a single RGB image. Our key insight is to exploit domain knowledge to regularize the network by deeply supervising its hidden layers, in order to sequentially infer intermediate concepts associated with the final task. To acquire training data in desired quantities with ground truth 3D shape and relevant concepts, we render 3D object CAD models to generate large-scale synthetic data and simulate challenging occlusion configurations between objects. We train the network only on synthetic data and demonstrate state-of-the-art performances on real image benchmarks including an extended version of KITTI, PASCAL VOC, PASCAL3D+ and IKEA for 2D and 3D keypoint localization and instance segmentation. The empirical results substantiate the utility of our deep supervision scheme by demonstrating effective transfer of knowledge from synthetic data to real images, resulting in less overfitting compared to standard end-to-end training.