Thanyaporn NoiplabMongkon SakdanupabAkara SupratakThanapong IntharahKhon Kaen UniversityMahidol UniversityTinyEpicBrains2020-01-272020-01-272019-09-01Proceedings - 2019 19th International Symposium on Communications and Information Technologies, ISCIT 2019. (2019), 470-4752-s2.0-85076221008https://repository.li.mahidol.ac.th/handle/20.500.14594/50615© 2019 IEEE. Currently, we are in an era when smartphones become everything even the best device to use for searching for a video from a video dataset using their camera. Designing an efficient dataset construction pipeline for the video retrieval system is not an obvious task and details of the process were often omitted in past literature. In this paper, we share our best practice of constructing a mobile video retrieval dataset. Every step from planning, developing mobile apps, recruiting workers, problems during and after the construction as well as the solutions are delineated in detail. In addition, we propose cloud architectures which facilitates the overall data collection process. This paper contributes to data science and computer vision research communities in two folds: first, it describes every step from planning to managing the dataset in detail, and second, it summarizes what should do and what should not: many of which are not obvious when planning for the dataset collection and during the data collection process.Mahidol UniversityComputer ScienceMathematicsConference PaperSCOPUS10.1109/ISCIT.2019.8905136