Conference paper Open Access

ITI-CERTH participation in ActEV and AVS Tracks of TRECVID 2022

Konstantinos Gkountakos; Damianos Galanopoulos; Despoina Touska; Konstantinos Ioannidis; Stefanos Vrochidis; Vasileios Mezaris; Ioannis Kompatsiaris

This report presents the overview of the runs related to Ad-hoc Video Search (AVS) and Activities in Extended Video (ActEV) tasks on behalf of the ITI-CERTH team. Our participation in the AVS task is based on a cross-modal deep network architecture utilizing several textual and visual features. As part of the retrieval stage, a dual-softmax approach is utilized to revise the calculated text-video
similarities. For the ActEV task, we adapt our framework to fit the new dataset and overcome the challenges of detecting and recognizing activities in a multi-label manner while experimenting with two separate activity classifiers.

Files (327.5 kB)
Name Size
327.5 kB Download
All versions This version
Views 3737
Downloads 1313
Data volume 4.3 MB4.3 MB
Unique views 2525
Unique downloads 77


Cite as