Lightweight 2D Map Construction of Vehicle Environments Using a Semi-Supervised Depth Estimation Approach
This paper addresses the problem of constructing a real-time 2D map for driving scenes from a single monocular RGB image. We presented a method based on three neural networks (depth estimation, 3D object detection, and semantic segmentation). We proposed a depth estimation neural network architectur...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-06-01
|
| Series: | Engineering Proceedings |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2673-4591/33/1/28 |
| _version_ | 1797381182500896768 |
|---|---|
| author | Alexey Kashevnik Ammar Ali |
| author_facet | Alexey Kashevnik Ammar Ali |
| author_sort | Alexey Kashevnik |
| collection | DOAJ |
| description | This paper addresses the problem of constructing a real-time 2D map for driving scenes from a single monocular RGB image. We presented a method based on three neural networks (depth estimation, 3D object detection, and semantic segmentation). We proposed a depth estimation neural network architecture that is fast and accurate in comparison with the state-of-the-art models. We designed our model to work in real time on light devices (such as an NVIDIA Jetson Nano and smartphones). The model is based on an encoder–decoder architecture with complex loss functions, i.e., normal loss, VNL, gradient loss (dx, dy), and mean absolute error. Our results show competitive results in comparison with the state-of-the-art methods, as our method is 30 times faster and smaller. |
| first_indexed | 2024-03-08T20:47:41Z |
| format | Article |
| id | doaj.art-49dc79f8e3c44510b7e3dbe19f119aff |
| institution | Directory Open Access Journal |
| issn | 2673-4591 |
| language | English |
| last_indexed | 2024-03-08T20:47:41Z |
| publishDate | 2023-06-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Engineering Proceedings |
| spelling | doaj.art-49dc79f8e3c44510b7e3dbe19f119aff2023-12-22T14:06:46ZengMDPI AGEngineering Proceedings2673-45912023-06-013312810.3390/engproc2023033028Lightweight 2D Map Construction of Vehicle Environments Using a Semi-Supervised Depth Estimation ApproachAlexey Kashevnik0Ammar Ali1SPC RAS, Saint Petersburg 199178, RussiaInformation Technology and Programming Faculry, ITMO University, Saint Petersburg 197101, RussiaThis paper addresses the problem of constructing a real-time 2D map for driving scenes from a single monocular RGB image. We presented a method based on three neural networks (depth estimation, 3D object detection, and semantic segmentation). We proposed a depth estimation neural network architecture that is fast and accurate in comparison with the state-of-the-art models. We designed our model to work in real time on light devices (such as an NVIDIA Jetson Nano and smartphones). The model is based on an encoder–decoder architecture with complex loss functions, i.e., normal loss, VNL, gradient loss (dx, dy), and mean absolute error. Our results show competitive results in comparison with the state-of-the-art methods, as our method is 30 times faster and smaller.https://www.mdpi.com/2673-4591/33/1/28machine learningdeep learningcomputer visionmonocular depth estimationreal-time depth estimationdriver assistant systems |
| spellingShingle | Alexey Kashevnik Ammar Ali Lightweight 2D Map Construction of Vehicle Environments Using a Semi-Supervised Depth Estimation Approach Engineering Proceedings machine learning deep learning computer vision monocular depth estimation real-time depth estimation driver assistant systems |
| title | Lightweight 2D Map Construction of Vehicle Environments Using a Semi-Supervised Depth Estimation Approach |
| title_full | Lightweight 2D Map Construction of Vehicle Environments Using a Semi-Supervised Depth Estimation Approach |
| title_fullStr | Lightweight 2D Map Construction of Vehicle Environments Using a Semi-Supervised Depth Estimation Approach |
| title_full_unstemmed | Lightweight 2D Map Construction of Vehicle Environments Using a Semi-Supervised Depth Estimation Approach |
| title_short | Lightweight 2D Map Construction of Vehicle Environments Using a Semi-Supervised Depth Estimation Approach |
| title_sort | lightweight 2d map construction of vehicle environments using a semi supervised depth estimation approach |
| topic | machine learning deep learning computer vision monocular depth estimation real-time depth estimation driver assistant systems |
| url | https://www.mdpi.com/2673-4591/33/1/28 |
| work_keys_str_mv | AT alexeykashevnik lightweight2dmapconstructionofvehicleenvironmentsusingasemisuperviseddepthestimationapproach AT ammarali lightweight2dmapconstructionofvehicleenvironmentsusingasemisuperviseddepthestimationapproach |