Multi-class recognition using noisy training data with a self-learning approach

Onderzoeksoutput: Bijdrage aan congresPaperAcademic

1 Downloads (Pure)

Samenvatting

Exploiting ConvNets for object classification systems requires extensive labor work, since these networks require to be trained by sufficiently large and accurately labeled datasets. We propose a novel self-learning approach, which is able to generate a reliable multi-class object classification model from a low-quality dataset that is disturbed with a high level of interclass noise samples. This approach iteratively purifies the noisy training datasets for each class and updates the classification model. The iterations continue until the model and its parameters reach sufficient quality. The self-learning approach based on ConvNets is evaluated for a maritime surveillance use case, where vessels need to be classified into eight different types. The experimental results on the evaluation dataset show that the proposed approach improves the F1 score approximately by 5%, 8% and 25% at the end of the third iteration, while the initial training datasets contain 40%, 50% and 60% interclass noise samples (erroneously classified labels of vessels), respectively. Additionally, the purification performance is highly dependent on inter- and inter-class similarities between training samples for higher noise levels. It was also found that the mean Average Precision (mAP) does not degrade so much, whereas other performance parameters show larger variation.
Originele taal-2Engels
Aantal pagina's8
DOI's
StatusGepubliceerd - 10 dec 2018
Evenement2018 Digital Image Computing: Techniques and Applications (DICTA) - Canberra, Australië
Duur: 10 dec 201813 dec 2018

Congres

Congres2018 Digital Image Computing: Techniques and Applications (DICTA)
LandAustralië
StadCanberra
Periode10/12/1813/12/18

Vingerafdruk Duik in de onderzoeksthema's van 'Multi-class recognition using noisy training data with a self-learning approach'. Samen vormen ze een unieke vingerafdruk.

  • Citeer dit

    Ghahremani, A., Bondarau, Y., & de With, P. H. N. (2018). Multi-class recognition using noisy training data with a self-learning approach. Paper gepresenteerd op 2018 Digital Image Computing: Techniques and Applications (DICTA), Canberra, Australië. https://doi.org/10.1109/DICTA.2018.8615864