ITI-CERTH participation in AVS Task of TRECVID 2023

This report presents an overview of the runs submitted to Ad-hoc Video Search (AVS) on behalf of the ITI-CERTH team. Our participation in the AVS task is based on a transformer-based extension of a cross-modal deep network architecture. We analyzed visual information at multiple levels of granularity using detected objects. During the retrieval stage, we employed a dual-softmax approach to adjust the calculated text-video similarities.