Sparse Representations for Object and Ego-motion Estimation in Dynamic Scenes
This addresses the problem of scale ambiguity and object motion in dynamic scenes for robotics and autonomous driving, representing an incremental improvement over existing methods.
The paper tackles the challenge of monocular visual odometry in dynamic scenes by proposing a learning-based method that predicts camera motion directly from optic flow, achieving state-of-the-art performance on trajectory and rotation prediction tasks on KITTI and Virtual KITTI datasets, with the sparse representation requiring only 5% of hidden neurons for best accuracy.
Dynamic scenes that contain both object motion and egomotion are a challenge for monocular visual odometry (VO). Another issue with monocular VO is the scale ambiguity, i.e. these methods cannot estimate scene depth and camera motion in real scale. Here, we propose a learning based approach to predict camera motion parameters directly from optic flow, by marginalizing depthmap variations and outliers. This is achieved by learning a sparse overcomplete basis set of egomotion in an autoencoder network, which is able to eliminate irrelevant components of optic flow for the task of camera parameter or motionfield estimation. The model is trained using a sparsity regularizer and a supervised egomotion loss, and achieves the state-of-the-art performances on trajectory prediction and camera rotation prediction tasks on KITTI and Virtual KITTI datasets, respectively. The sparse latent space egomotion representation learned by the model is robust and requires only 5% of the hidden layer neurons to maintain the best trajectory prediction accuracy on KITTI dataset. Additionally, in presence of depth information, the proposed method demonstrates faithful object velocity prediction for wide range of object sizes and speeds by global compensation of predicted egomotion and a divisive normalization procedure.