10.1007/978-3-319-51811-4_9
https://zenodo.org/records/240853
oai:zenodo.org:240853
Pittaras, Nikiforos
Nikiforos
Pittaras
Information Technologies Institute (ITI), Centre for Research and Technology Hellas (CERTH), Thessaloniki, Greece
Markatopoulou, Foteini
Foteini
Markatopoulou
Information Technologies Institute (ITI), Centre for Research and Technology Hellas (CERTH), Thessaloniki, Greece
Mezaris, Vasileios
Vasileios
Mezaris
Information Technologies Institute (ITI), Centre for Research and Technology Hellas (CERTH), Thessaloniki, Greece
Patras, Ioannis
Ioannis
Patras
Queen Mary University of London, Mile end Campus, UK
Comparison of Fine-tuning and Extension Strategies for Deep Convolutional Neural Networks
Zenodo
2016
2016-12-31
https://zenodo.org/communities/invid-h2020
https://zenodo.org/communities/eu
Creative Commons Attribution 4.0 International
In this study we compare three different fine-tuning strategies in order to investigate the best way to transfer the parameters of popular deep convolutional neural networks that were trained for a visual annotation task on one dataset, to a new, considerably different dataset. We focus on the concept-based image/video annotation problem and use ImageNet as the source dataset, while the TRECVID SIN 2013 and PASCAL VOC-2012 classification datasets are used as the target datasets. A large set of experiments examines the effectiveness of three fine-tuning strategies on each of three different pre-trained DCNNs and each target dataset. The reported results give rise to guidelines for effectively fine-tuning a DCNN for concept-based visual annotation.
European Commission
10.13039/501100000780
687786
In Video Veritas – Verification of Social Media Video Content for the News Industry