Mono3D++: Monocular 3D Vehicle Detection with Two-Scale 3D Hypotheses and Task Priors
This work addresses the challenge of 3D vehicle detection from monocular images, which is important for autonomous driving and robotics, but it appears incremental as it builds on existing morphable models and priors.
The paper tackles the ill-posed problem of inferring 3D pose and shape of vehicles from a single image by optimizing two-scale projection consistency and integrating task priors, resulting in improved robustness and accuracy in monocular 3D vehicle detection.
We present a method to infer 3D pose and shape of vehicles from a single image. To tackle this ill-posed problem, we optimize two-scale projection consistency between the generated 3D hypotheses and their 2D pseudo-measurements. Specifically, we use a morphable wireframe model to generate a fine-scaled representation of vehicle shape and pose. To reduce its sensitivity to 2D landmarks, we jointly model the 3D bounding box as a coarse representation which improves robustness. We also integrate three task priors, including unsupervised monocular depth, a ground plane constraint as well as vehicle shape priors, with forward projection errors into an overall energy function.