Cross-Modal Knowledge Distillation for Depth Privileged Monocular Visual Odometry

Li, Bin; Wang, Shuling; Ye, Haifeng; Gong, Xiaojin; Xiang, Zhiyu

doi:10.1109/lra.2022.3166457

Cited by 6 publications

(2 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The authors in [24] further proposed to utilize autoencoder networks to learn a better representations for their optical flow maps. To eliminate the noises from input observations, a number of researchers have also investigated attention based VO methods [21,26,27,28,32,36,37,38] and uncertainty based VO methods [22,41,44,45] to mitigate the impacts of moving objects from their input observations.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Pixel-Wise Prediction based Visual Odometry via Uncertainty Estimation

Chen¹,

Liao²,

Yang³

et al. 2022

Preprint

View full text Add to dashboard Cite

This paper introduces pixel-wise prediction based visual odometry (PWVO), which is a dense prediction task that evaluates the values of translation and rotation for every pixel in its input observations. PWVO employs uncertainty estimation to identify the noisy regions in the input observations, and adopts a selection mechanism to integrate pixel-wise predictions based on the estimated uncertainty maps to derive the final translation and rotation. In order to train PWVO in a comprehensive fashion, we further develop a data generation workflow for generating synthetic training data. The experimental results show that PWVO is able to deliver favorable results. In addition, our analyses validate the effectiveness of the designs adopted in PWVO, and demonstrate that the uncertainty maps estimated by PWVO is capable of capturing the noises in its input observations.

show abstract

Section: Related Workmentioning

confidence: 99%

“…[32,33,34,35]. Attention mechanisms seek the clues of possible candidates from intermediate representations [21,26,27,28,29,32,36,37,38,39]. Uncertainty estimation, on the other hand, implicitly captures the noises and enables the models to respond to the measured stochasticity inherent in the observations [11,22,40,41,42,43,44,45].…”

Section: Introductionmentioning

confidence: 99%