Description:

1. manual_icp/sift_icp.cpp : Gives the R-T and the loop closing edge for g2o file, RGB and Depth images are used to calculate SIFT features and the correspondences. Then the depth image is converted into PointCloud which is given to the ICP along with correspondences.

   1. Usage: ./sift_icp rgb1.png rgb2.png depth1.png depth2.png
   2. Download data: from here

2. Reading Resources :

   1. ICP
   2. Concatenation of 2 cloud
   3. Depth2pointcloud
   4. OpenCV feature detection
   5. Manual correspondences function
   6. Pixel coordinates from key point

3. cloud_assemble/cloud_assemble.cpp : Concatanates the PointClouds on the basis of optimized g2o file. Load the first PointCloud and add it to the global map, then rest PointClouds are individually transformed according to the optimized edge constraints in "EDGE_SE3:QUAT" and added to global map to form a single pcd file. While the individual cloud is also colored after loading based on its topology predicted by the network.

   1. Usage: ./cloud_assemble file.g2o /path/to/clouds/ images_file.txt predictions_file.txt
   2. Download data: from the above mentioned database download link

4. launch_scripts : Contains launching scripts and configurations files for RTABMAP and Realsense.

5. docs/: Contains reading resources to papers.

6. First warehouse results : Download dataset from here