当前位置:网站首页>Seasonal depth: cross seasonal monocular depth prediction dataset and benchmark (CS CV) in multiple environments

Seasonal depth: cross seasonal monocular depth prediction dataset and benchmark (CS CV) in multiple environments

2020-12-07 19:23:49 Ling Qian

Monocular depth prediction is a hot topic in recent years , But because of the lack of real-world data sets and benchmarks , There are few studies on depth prediction in various environments such as illumination and seasonal variation . In this work , We from CMU A new cross season scale-free monocular depth prediction dataset based on structure and motion is proposed . Then we developed several metrics to measure performance in different environments , Use the latest open source deep prediction pre training model from KITTI Benchmark of data set . Through extensive zero emission experiments on the proposed data sets , We show that long-term monocular depth prediction is far from solved , And offers promising solutions , In the future work , Including geometry or scale invariant training . Besides , Multi environment synthetic data set and cross data set verification are beneficial to the robustness of real environment variance .

Original title :SeasonDepth: Cross-Season Monocular Depth Prediction Dataset and Benchmark under Multiple Environments

original text :Monocular depth prediction has been well studied recently, while there are few works focused on the depth prediction across multiple environments, e.g. changing illumination and seasons, owing to the lack of such real-world dataset and benchmark. In this work, we derive a new cross-season scaleless monocular depth prediction dataset SeasonDepth from CMU Visual Localization dataset through structure from motion. And then we formulate several metrics to benchmark the performance under different environments using recent stateof-the-art open-source depth prediction pretrained models from KITTI benchmark. Through extensive zero-shot experimental evaluation on the proposed dataset, we show that the long-term monocular depth prediction is far from solved and provide promising solutions in the future work, including geometricbased or scale-invariant training. Moreover, multi-environment synthetic dataset and cross-dataset validataion are beneficial to the robustness to real-world environmental variance.

Original author :Hanjiang Hu, Baoquan Yang, Weiang Shi, Zhijian Qiao, Hesheng Wang

Original address :https://arxiv.org/abs/2011.04408

Original statement , This article is authorized by the author + Community publication , Unauthorized , Shall not be reproduced .

If there is any infringement , Please contact the yunjia_community@tencent.com Delete .

版权声明
本文为[Ling Qian]所创,转载请带上原文链接,感谢
https://chowdera.com/2020/11/20201119031742516f.html