The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLIII-B2-2020
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B2-2020, 583–588, 2020
https://doi.org/10.5194/isprs-archives-XLIII-B2-2020-583-2020
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B2-2020, 583–588, 2020
https://doi.org/10.5194/isprs-archives-XLIII-B2-2020-583-2020

  12 Aug 2020

12 Aug 2020

KNOWLEDGE DISTILLATION USING GANS FOR FAST OBJECT DETECTION

E. Finogeev1,2, V. Gorbatsevich1, A. Moiseenko1, Y. Vizilter1, and O. Vygolov1 E. Finogeev et al.
  • 1Federal State Unitary Enterprise «State Research Institute of Aviation Systems», Russian Federation
  • 2National Research Nuclear University “MEPhI”, Russian Federation

Keywords: CNN, Object detection, Knowledge distillation, Real-time, GAN, Single shot detector

Abstract. In this paper, we propose a new method for knowledge distilling based on generative adversarial networks. Discriminator CNNs is used as an adaptive knowledge distilling loss. In experiments, single shot multibox detector SSD based on MobileNet v2 and ShuffleNet v1 are used as student networks. Our tests showed AP and mAP improvement of more than 3% on PascalVOC and 1% on MS Coco datasets compared with the baseline algorithm without any architecture or dataset changes. The proposed approach is general and can be used not only with SSD but also with any type of object detection algorithms.