Nicola Fioraio - PhD thesis
In this page we provide a direct link to Nicola Fioraio's PhD thesis and additional results referenced in the text.
Thesis
Full text [PDF]
Additional Material
Chapter 2
RGB-D SLAM For Mobile Devices
In Chap. 2 we describe SlamDunk, a scalable and effective solution for RGB-D SLAM. Here we demonstrate live reconstruction for both the desktop (Sec. 2.3) and the mobile (Sec. 2.4) versions. Also, we show object and room reconstruction performed with an Asus Xtion PRO Live sensor.
SlamDunk (desktop)
Reconstruction of a chair (cfr. middle image in Fig. 2.0.1) [MP4]
Reconstruction of a room (cfr. Fig. 2.5.2) [MP4]
Live reconstruction and recovery after occlusion [AVI]
SlamDunk (android)
Live reconstruction of a kitchen (cfr. Fig. 2.5.3b) [WMV]
Live reconstruction of a room [WMV]
Chapter 4
Semantic Bundle Adjustment
In Chap. 4 we investigate joint object detection and SLAM. The supplementary material produced for our CVPR 2013 paper (N. Fioraio, L. Di Stefano, "Joint Detection, Tracking and Mapping by Semantic Bundle Adjustment") compares the reconstruction for both the 4-objects and 7-objects sequences (Sec. 4.4) to the plain SLAM approach, highlighting the effect of the validation graph on object detection and global alignment (cfr. Tab. 4.4.1). Also, we show results on real data (cfr. Fig. 4.4.5) and object-aware augmented reality with occlusion handling (cfr. Fig. 4.4.8)
Supplementary material submitted with: N. Fioraio, L. Di Stefano, "Joint Detection, Tracking and Mapping by Semantic Bundle Adjustment", Computer Vision and Pattern Recognition (CVPR), 2013. [AVI]