vision_3d
=========

ROS Stack used for 3D vslam and object detection algorithms.
I will add object recognition algorithms which use RGB-D information for use with robotics.

Pre-requisites
=========
opencv2 (tested with 2.4.2, comes with ros-fuerte-desktop-full)
pcl (tested with 1.5.1, comes with ros-fuerte-desktop-full)
openni_launch ROS Stack for fuerte which includes OpenNI driver (comes with ros-fuerte-desktop-full)
pcl_ros (comes with ros-fuerte-desktop-full)

TODO
=========

Feature Extraction:

	-Implement SURF3D descriptor
	-Implement SIFT3D detector and descriptor
	-Implement SURFLET descriptor
	-Implement Kernel descriptors

MISC:

	-Add functionality to convert RGB-D images to point clouds and vice-versa

Saliency Maps:

	-Research and implement 2D and 3D saliency maps for attention focusing
	-Maybe use saliency_detector ROS STACK
	-3D cues with IKN Maps
	-3D saliency maps by Pirri et al
	

Classifiers:

	-Implement SVM classifier for training (using libSVM, probably) with different kernels
	-Implement adaboost classifier
	-Experiment with part-based object detection

Object detection:

	-Implement node for detecting certain types of objects using RGB-D data from kinect