Towards Spatial Perception: Learning to Locate Objects from Vision #robotDoC
1 like858 views
This document summarizes Jürgen Leitner's work on spatial perception and object localization using computer vision. It discusses detecting objects in images, developing an icVision framework, current localization approaches, and using learning methods like neural networks and genetic programming to locate objects spatially. The goal is to train robots to perceive spatial relationships and continuously learn object locations from visual input.
Towards Spatial Perception: Learning to Locate Objects from Vision #robotDoC
- 1. Jürgen ’Juxi’ Leitner S. Harding, M. Frank, A. Förster, J. Schmidhuber istituto dalle molle di studi sull’intelligenza artificiale università della svizzera italiana idsia / usi / supsi towards spatial perception learning to locate objects from vision #robotdoc 2012 Monday, September 10, 2012
- 2. iCub thanks to G. Metta, IIT and U.Plymouth for their pictures Monday, September 10, 2012
- 3. our iCub setup is for manipulation Monday, September 10, 2012
- 4. thanks to G. Metta and IIT for this picture visual perception Monday, September 10, 2012
- 5. three parts Monday, September 10, 2012
- 6. detecting objects Harding et al., GPTP 2012, Leitner et al., ICDL 2012 Monday, September 10, 2012
- 7. Image credit: NASA also works for detecting rocks on Mars Leitner et al., iSAIRAS 2012 Monday, September 10, 2012
- 8. detecting objects Harding et al., GPTP 2012, Leitner et al., ICDL 2012 Monday, September 10, 2012
- 9. icVision framework Leitner et al., BICA 2012 Monday, September 10, 2012
- 10. three parts GPTP2012, ICDL2012(?) BICA 2012 Monday, September 10, 2012
- 11. locate objects Monday, September 10, 2012
- 12. current localisation approaches Monday, September 10, 2012
- 13. learning approach + Monday, September 10, 2012
- 14. learning setup Monday, September 10, 2012
- 15. spatial perception neural network Monday, September 10, 2012
- 16. spatial perception genetic programming Monday, September 10, 2012
- 17. x = 17.81 − 0.01906 v1 + 0.1527 v4 + 0.1378 v7 + 0.01108 v10 − 0.0296 v11 − 0.1207 v12 y = 1.124224045 + 0.1295920897 v10 + 0.1156011386 v8 + 0.01695234993 v0 spatial perception genetic programming Monday, September 10, 2012
- 18. relative errors Monday, September 10, 2012
- 19. transferring spatial perception Monday, September 10, 2012
- 20. future work automate the learning (IROS 2012) different spatial representations “intrinsic motivation” learning online, continuous learning eye-hand coordination Monday, September 10, 2012
- 21. thanks for listening juxi@idsia.ch http://Juxi.net/projects Monday, September 10, 2012