Realistic images generation for training artificial neural networks in robot navigation problem
Abstract
The problem of obtaining training dataset for setting weights of neural network designed for indoor detection of doors is considered and solved on the particular example. A method for generating realistic synthetic data is developed. The method involves replacing a priori known target objects on digital images with new reference objects that were obtained by projective transformation of reference objects. The method is designed to obtain training dataset for training and testing of artificial neural networks, which will be used in the mobile robot control system to solve autonomous navigation problem. The effectiveness of the proposed method was confirmed experimentally.
About the Author
L. A. Khodasevich
The United Institute of Informatics Problems, National Academy of Sciences of Belarus
Belarus
Belarus
Liubov A. Khodasevich - Trainee of Junior Researcher, The United Institute of Informatics Problems of the National Academy of Sciences of Belarus
6, Surganova Str., 220012, Minsk
References
1. LeCun Y., Jarrett K., Kavukcuoglu K., Ranzato M. What is the best multi-stage architecture for object recognition? The 12th International Conference on Computer Vision, Kyoto, 27 September - 4 October 2009. Kyoto, 2009, pp. 2146-2153.
2. Kortylewski A., Schneider A., Gerig T., Egger B., Morel-Foster A., Vetter T. Training deep face recognition systems with synthetic data. Available at: https://arxiv.org/pdf/1802.05891.pdf (accessed 16.04.2018).
3. Chigorin A., Moiseev B. Klassifikatsiya avtodorozhnykh znakov na osnove svertochnoi neiroseti, obuchennoi na sinteticheskikh dannykh [Road signs classification based on convolutional neural network that trained on synthetic data]. Trudy 22 Mezhdunarodnoj konferencii po komp'jutemoj grafike i mashinnomu zreniju GrafiKon’2012 [Proceedings of the 22nd International Conference on Computer Graphics and Computer Vision GraphiCon 2012, Moscow, 1-5 October 2012]. Moscow, 2012, pp. 284-287 (in Russian).
4. Prokopovich G. A. Razrabotka sistemy tehnicheskogo zrenija dlja servisnogo mobil'nogo robota [Development of technical vision system for service mobile robot]. Tretiy Vserossiyskiy naychno-prakticheskiy seminar "Bespilotnye transportnye sredstva s elementami isskustvennogo intellekta" [The Third All-Russian Scientific and Practical Seminar "Unmanned Vehicles with Artificial Intelligence Elements", Innopolis, 22-23 September 2015]. Innopolis, Universitet Innopolis, 2016, pp. 127-136 (in Russian).
5. Uijlings J. R. R., Sande K. E. A., Gevers T., Smeulders A. W. M. Selective search for object recognition. Available at: http://www.huppelen.nl/publications/selectiveSearchDraft.pdf (accessed 28.04.2018).
6. Ren S., He K., Girshick R., Sun J. Faster R-CNN: Towards real-time object detection with region proposal networks. Available at: https://arxiv.org/pdf/1506.01497.pdf (accessed 15.05.2018).
7. Szegedy C., Vanhoucke V., Ioffe S., Shlens J. Rethinking the inception architecture for computer vision. Available at: https://arxiv.org/pdf/1512.00567.pdf (accessed 15.05.2018).
8. LeCun Y., Cortes C., Burges C. J. C. The MNIST database. Available at: http://yann.lecun.com/exdb/mnist/ (accessed 16.03.2018).
9. Lin T., Patterson G., Ronchi M. R., Cui Y., Dollar P. Microsoft COCO: Common objects in context. Available at: http://cocodataset.org/#home (accessed 16.03.2018).
10. Tzutalin /labelImg. Available at: https://github.com/tzutalin/labelImg (accessed 18.06.2018).
Review
For citations:
Khodasevich L.A. Realistic images generation for training artificial neural networks in robot navigation problem. Informatics. 2018;15(4):50-58. (In Russ.)
Views: 792
Email this article 