A Recurrent Deep Network for Estimating the Pose of Real Indoor Images from Synthetic Image Sequences
Recently, deep convolutional neural networks (CNN) have become popular for indoor visual localisation, where the networks learn to regress the camera pose from images directly. However, these approaches perform a 3D image-based reconstruction of the indoor spaces beforehand to determine camera poses...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-09-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/20/19/5492 |
| _version_ | 1797552606038458368 |
|---|---|
| author | Debaditya Acharya Sesa Singha Roy Kourosh Khoshelham Stephan Winter |
| author_facet | Debaditya Acharya Sesa Singha Roy Kourosh Khoshelham Stephan Winter |
| author_sort | Debaditya Acharya |
| collection | DOAJ |
| description | Recently, deep convolutional neural networks (CNN) have become popular for indoor visual localisation, where the networks learn to regress the camera pose from images directly. However, these approaches perform a 3D image-based reconstruction of the indoor spaces beforehand to determine camera poses, which is a challenge for large indoor spaces. Synthetic images derived from 3D indoor models have been used to eliminate the requirement of 3D reconstruction. A limitation of the approach is the low accuracy that occurs as a result of estimating the pose of each image frame independently. In this article, a visual localisation approach is proposed that exploits the spatio-temporal information from synthetic image sequences to improve localisation accuracy. A deep Bayesian recurrent CNN is fine-tuned using synthetic image sequences obtained from a building information model (BIM) to regress the pose of real image sequences. The results of the experiments indicate that the proposed approach estimates a smoother trajectory with smaller inter-frame error as compared to existing methods. The achievable accuracy with the proposed approach is 1.6 m, which is an improvement of approximately thirty per cent compared to the existing approaches. A Keras implementation can be found in our Github repository. |
| first_indexed | 2024-03-10T16:03:29Z |
| format | Article |
| id | doaj.art-39154f4832484f3a921b58351622c599 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-10T16:03:29Z |
| publishDate | 2020-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-39154f4832484f3a921b58351622c5992023-11-20T15:03:19ZengMDPI AGSensors1424-82202020-09-012019549210.3390/s20195492A Recurrent Deep Network for Estimating the Pose of Real Indoor Images from Synthetic Image SequencesDebaditya Acharya0Sesa Singha Roy1Kourosh Khoshelham2Stephan Winter3Department of Infrastructure Engineering, The University of Melbourne, Parkville, Victoria 3010, AustraliaInstitute for Sustainable Industries and Livable Cities, Victoria University, Werribee, Victoria 3030, AustraliaDepartment of Infrastructure Engineering, The University of Melbourne, Parkville, Victoria 3010, AustraliaDepartment of Infrastructure Engineering, The University of Melbourne, Parkville, Victoria 3010, AustraliaRecently, deep convolutional neural networks (CNN) have become popular for indoor visual localisation, where the networks learn to regress the camera pose from images directly. However, these approaches perform a 3D image-based reconstruction of the indoor spaces beforehand to determine camera poses, which is a challenge for large indoor spaces. Synthetic images derived from 3D indoor models have been used to eliminate the requirement of 3D reconstruction. A limitation of the approach is the low accuracy that occurs as a result of estimating the pose of each image frame independently. In this article, a visual localisation approach is proposed that exploits the spatio-temporal information from synthetic image sequences to improve localisation accuracy. A deep Bayesian recurrent CNN is fine-tuned using synthetic image sequences obtained from a building information model (BIM) to regress the pose of real image sequences. The results of the experiments indicate that the proposed approach estimates a smoother trajectory with smaller inter-frame error as compared to existing methods. The achievable accuracy with the proposed approach is 1.6 m, which is an improvement of approximately thirty per cent compared to the existing approaches. A Keras implementation can be found in our Github repository.https://www.mdpi.com/1424-8220/20/19/5492indoor localisationcamera pose regression3D building modelslong short term memory |
| spellingShingle | Debaditya Acharya Sesa Singha Roy Kourosh Khoshelham Stephan Winter A Recurrent Deep Network for Estimating the Pose of Real Indoor Images from Synthetic Image Sequences Sensors indoor localisation camera pose regression 3D building models long short term memory |
| title | A Recurrent Deep Network for Estimating the Pose of Real Indoor Images from Synthetic Image Sequences |
| title_full | A Recurrent Deep Network for Estimating the Pose of Real Indoor Images from Synthetic Image Sequences |
| title_fullStr | A Recurrent Deep Network for Estimating the Pose of Real Indoor Images from Synthetic Image Sequences |
| title_full_unstemmed | A Recurrent Deep Network for Estimating the Pose of Real Indoor Images from Synthetic Image Sequences |
| title_short | A Recurrent Deep Network for Estimating the Pose of Real Indoor Images from Synthetic Image Sequences |
| title_sort | recurrent deep network for estimating the pose of real indoor images from synthetic image sequences |
| topic | indoor localisation camera pose regression 3D building models long short term memory |
| url | https://www.mdpi.com/1424-8220/20/19/5492 |
| work_keys_str_mv | AT debadityaacharya arecurrentdeepnetworkforestimatingtheposeofrealindoorimagesfromsyntheticimagesequences AT sesasingharoy arecurrentdeepnetworkforestimatingtheposeofrealindoorimagesfromsyntheticimagesequences AT kouroshkhoshelham arecurrentdeepnetworkforestimatingtheposeofrealindoorimagesfromsyntheticimagesequences AT stephanwinter arecurrentdeepnetworkforestimatingtheposeofrealindoorimagesfromsyntheticimagesequences AT debadityaacharya recurrentdeepnetworkforestimatingtheposeofrealindoorimagesfromsyntheticimagesequences AT sesasingharoy recurrentdeepnetworkforestimatingtheposeofrealindoorimagesfromsyntheticimagesequences AT kouroshkhoshelham recurrentdeepnetworkforestimatingtheposeofrealindoorimagesfromsyntheticimagesequences AT stephanwinter recurrentdeepnetworkforestimatingtheposeofrealindoorimagesfromsyntheticimagesequences |