HKUST-Aerial-Robotics / DenseSurfelMapping

Thank you for your work，and release it. When i use ORB_slam2 model RGBd
to run Kitti dataset, sequence 00, there is no problem. And i can get the ideal result. But i processing our outdoor road data, it don't extract the super pixels, and don't generate the point cloud. I don't know which problem it is. Can you help me to solve it? I am looking forward your reply!!! Thank you!

I have the same problem,and the spimage also can not received when using mynteye camera,but the surfel_color02.bag is right.so whether need to modify some code to solve these problems, except for vins_realsense.launch.looking forward for your reply.thanks~ @dvorak0 @WayneTimer @groundmelon @shaojie

Hi~~,At present, I can extract super pixels. When reading depth map, pay attention to the unit of depth image is meter. After changing the information of reading depth map in ' publisher.py ' , I can get super pixels, but the reconstruction effect is very poor, running orbslam2 rgbd mode , Core dump was appeared. Because the quliaty of depth image is poor? Hope to provide some ideas and solutions. Thanks!!! I am looking forward to your replay!

Successfully running the program is good news, although the effect is not very good. Can you tell me where I need to modify pubilsher.py? Are there other codes that need to be modified? I am using a stereo camera, not an rgb-d camera. I cannot run this program normally at the moment, Thanks in advance. Do you have e-mail or wechat or qq,It is more convenient to communicate with chat tools.
@PumpkinRun

Hi, I am here for help.
@PumpkinRun can you convert the depth maps into point clouds and show me? From the normal map, it seems quite noisy.

@Shame-fight The messages should be well synchronized. If you do not receive any messages please go to

DenseSurfelMapping/surfel_fusion/src/surfel_map.cpp

Line 83 in 77f65b1

void SurfelMap::image_input(const sensor_msgs::ImageConstPtr &image_input)

,

DenseSurfelMapping/surfel_fusion/src/surfel_map.cpp

Line 103 in 77f65b1

void SurfelMap::synchronize_msgs()

,

DenseSurfelMapping/surfel_fusion/src/surfel_map.cpp

Line 93 in 77f65b1

void SurfelMap::depth_input(const sensor_msgs::ImageConstPtr &depth_input)

, and

DenseSurfelMapping/surfel_fusion/src/surfel_map.cpp

Line 205 in 77f65b1

void SurfelMap::orb_results_input(

for debugging.
I do not know much about your senesors and configuration, so please refer to the example bag and configure your settings.

Hi, I am here for help.
@PumpkinRun can you convert the depth maps into point clouds and show me? From the normal map, it seems quite noisy.
Hello, Thank you for your reply. I had converted the depth maps into point clouds and discoveried it is terrible.

This is kitti 00 sequence image00 , 000000.png and 000000.npy.

This is owner data including image_raw and point clouds.

So, It should be the problem of depth map, which makes the reconstruction result poor

Hi, @PumpkinRun not sure what's on the ground. It does not seems right, right?

NO superpixel norm