Visual Computing

University of Konstanz
Proceedings of the 7th Annual ACM Workshop on the Lifelog Search Challenge,

Searching Temporally Distant Activities in Lifelog Data With PraK Tool V2

Z. Vopálková, J. Yaghob, M. Stroh, U. Schlegel, J. Lokoc
Teaser of Searching Temporally Distant Activities in Lifelog Data With PraK Tool V2

Architecture of the presented video search system based on stateless Video Data Services. The Application Backend handles Bayesian re-ranking and state for various users, whileWeb Frontend allows interaction with the system.

Material

Paper (.pdf, 1.6MB)

Abstract

We present a new version of our interactive cross-modal search system PraK that was updated for the needs of lifelog search challenge data and tasks. The architecture of PraK allows to hide lifelog data specific details behind already existing interface of video data services. Hence, the already existing application service designed for the Video Browser Showdown could be tested for the visual part of the lifelog dataset. Nevertheless, we add more features to address search challenges specific for lifelog data. In the new version, the interface of the services is slightly updated to incorporate meta-data based filtering as well as to support temporally distant activity queries. Another major update of the system is shift of state management from the web frontend to the application backend, which leads to more centralized architecture but allows to deploy PraK components in a more efficient way.

BibTeX

@inproceedings{Vopalkova2024SearchingTemporallyDistant,
  abstract  = {We present a new version of our interactive cross-modal search system PraK that was updated for the needs of lifelog search challenge data and tasks. The architecture of PraK allows to hide lifelog data specific details behind already existing interface of video data services. Hence, the already existing application service designed for the Video Browser Showdown could be tested for the visual part of the lifelog dataset. Nevertheless, we add more features to address search challenges specific for lifelog data. In the new version, the interface of the services is slightly updated to incorporate meta-data based filtering as well as to support temporally distant activity queries. Another major update of the system is shift of state management from the web frontend to the application backend, which leads to more centralized architecture but allows to deploy PraK components in a more efficient way.},
  address   = {New York, NY, USA},
  author    = {Z. Vopálková, J. Yaghob, M. Stroh, U. Schlegel, J. Lokoc},
  booktitle = {Proceedings of the 7th Annual ACM Workshop on the Lifelog Search Challenge},
  doi       = {10.1145/3643489.3661131},
  isbn      = {9798400705502},
  keywords  = {lifelog data, interactive retrieval systems, similarity search},
  location  = {Phuket, Thailand},
  numpages  = {6},
  pages     = {111–116},
  publisher = {Association for Computing Machinery},
  series    = {LSC '24},
  title     = {Searching Temporally Distant Activities in Lifelog Data With PraK Tool V2},
  url       = {https://doi.org/10.1145/3643489.3661131},
  year      = {2024}
}