Frantic race for higher-quality images, at what cost? - Insep - Institut national du sport, de l'expertise et de la performance Accéder directement au contenu
Poster Année : 2022

Frantic race for higher-quality images, at what cost?

Course effrénée aux images de meilleure qualité, à quel prix ?

Résumé

When implementing computer vision algorithms, the quality of the video input is a crucial parameter for systematic application as : • Training and inference conditions of Deep Learning models may differ. • Video quality may be downgraded due to sourcing or transfer encoding. • Storage capacity may limit the volume/quantity of data available. To better understand the sensitivity of inference with regarding video quality, we investigated the effect of video compression on the output of a human pose estimation network model.
Fichier principal
Vignette du fichier
poster_IDESSAI_schortgen.pdf (1.48 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04032785 , version 1 (16-03-2023)

Identifiants

  • HAL Id : hal-04032785 , version 1

Citer

Alexandre Schortgen, Lionel Reveret, Guillaume Saulière, Antoine Muller, Thibault Goyallon, et al.. Frantic race for higher-quality images, at what cost?: Application of computer vision models to high level boxing test matches. 2nd Inria-DFKI European Summer School on AI (IDESSAI 2022), Aug 2022, Saarbrücken, France. ⟨hal-04032785⟩
63 Consultations
16 Téléchargements

Partager

Gmail Facebook X LinkedIn More