Open Access Dynamic Human Point Cloud Datasets
Pitkänen, Mikko (2023)
Pitkänen, Mikko
2023
Tieto- ja sähkötekniikan kandidaattiohjelma - Bachelor's Programme in Computing and Electrical Engineering
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2023-08-08
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202306146735
https://urn.fi/URN:NBN:fi:tuni-202306146735
Tiivistelmä
Video conferencing tools in use today transmit 2-dimensional (2D) video. 2D video lacks the depth dimension, meaning the distance between an object and the camera. Therefore, 2D video may not be viewed from other angles in order to see behind captured objects. A more immersive form of tele-communication, known as tele-presence, instead utilizes volumetric video. Volumetric video is 3-dimensional (3D) video that can be viewed from any angle. Tele-presence systems create virtual spaces where the users are able to interact with the environment and with each other like in the real world.
Volumetric video may be implemented using point clouds. Point clouds are sets of sampled data points that represent surfaces in 3D space. However, point clouds require large amounts of storage. In order to encode point clouds into a less memory intensive format, point cloud compression is required. The development of tele-presence and point cloud compression technologies is currently lagging due to a lack of diverse test data.
This thesis evaluates the state of currently available point cloud datasets and their usability in development of the technologies mentioned above and compares the datasets. In order to ensure that the compared datasets match the planned use cases, a set of selection criteria is defined. In addition to the subjective visual quality of each dataset, the number of point clouds within each dataset and their contents are also considered.
As a result of the comparison, it is found that video produced using point clouds has lower visual quality than video produced using the commonly used textured meshes. The largest of the datasets, CWIPC-SXR, is found to be a versatile test dataset for tele-presence, due to the nature of the social interactions depicted in it. There still exists a very limited number of datasets matching the specified selection criteria and all except one contain only 1-4 point clouds. Therefore, more point cloud datasets meeting the criteria must be captured.
Volumetric video may be implemented using point clouds. Point clouds are sets of sampled data points that represent surfaces in 3D space. However, point clouds require large amounts of storage. In order to encode point clouds into a less memory intensive format, point cloud compression is required. The development of tele-presence and point cloud compression technologies is currently lagging due to a lack of diverse test data.
This thesis evaluates the state of currently available point cloud datasets and their usability in development of the technologies mentioned above and compares the datasets. In order to ensure that the compared datasets match the planned use cases, a set of selection criteria is defined. In addition to the subjective visual quality of each dataset, the number of point clouds within each dataset and their contents are also considered.
As a result of the comparison, it is found that video produced using point clouds has lower visual quality than video produced using the commonly used textured meshes. The largest of the datasets, CWIPC-SXR, is found to be a versatile test dataset for tele-presence, due to the nature of the social interactions depicted in it. There still exists a very limited number of datasets matching the specified selection criteria and all except one contain only 1-4 point clouds. Therefore, more point cloud datasets meeting the criteria must be captured.
Kokoelmat
- Kandidaatintutkielmat [8798]