search

ravich2-7183 / openDTAM

An opensource implementation of DTAM
6 stars 5 forks source link

You use the keyframe and pose of keyframe to reconstruction? #2

Open nonlinear1 opened 4 years ago

nonlinear1 commented 4 years ago

I am change the openDTAM to let dense reconstruction from keyframe and pose of keyframe. So I see your codes has realized this?

ravich2-7183 commented 4 years ago

Yes, the dense reconstruction was implemented.

nonlinear1 commented 4 years ago

Thank you very much for your good work!

ravich2-7183 commented 4 years ago

Thank you very much. I want your code output the results of reconstruction in ply or pcd format. So could you give me some suggestions?

Right now, I'm not actively working on openDTAM. I might look at this more closely on the weekend. In the meantime see if the following branch helps:

https://github.com/ravich2-7183/openDTAM/tree/publish_depth/src/dense-mapping

Also note the following points:

  1. My version (ros-openDTAM) is based on ROS-Indigo and Ubuntu-14.04. I failed to build it using ROS-kinetic and Ubuntu-16.04, because ROS-kinetic incorporated openCV-3 into ROS, while ros-openDTAM along with ORB_SLAM and InfiniTAM-ros all depend on openCV-2.4.9, .

  2. While all the pieces required for a complete system are there, a thorough systems integration and testing was not completed to ensure that adequate dense-mapping quality is obtained in common scenarios. Most of the testing was done against a few minute long ros-bag-files.

nonlinear1 commented 4 years ago

Thank you very much!

At 2020-01-08 01:06:09, "Engr. Ravi" notifications@github.com wrote:

Thank you very much. I want your code output the results of reconstruction in ply or pcd format. So could you give me some suggestions?

Right now, I'm not actively working on openDTAM. I might look at this more closely on the weekend. In the meantime see if the following branch helps:

https://github.com/ravich2-7183/openDTAM/tree/publish_depth/src/dense-mapping

Also note the following points:

My version (ros-openDTAM) is based on ROS-Indigo and Ubuntu-14.04. I failed to build it using ROS-kinetic and Ubuntu-16.04, because ROS-kinetic incorporated openCV-3 into ROS, while ros-openDTAM along with ORB_SLAM and InfiniTAM-ros all depend on openCV-2.4.9, .

While all the pieces required for a complete system are there, a thorough systems integration and testing was not completed to ensure that adequate dense-mapping quality is obtained in common scenarios. Most of the testing was done against a few minute long ros-bag-files.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

nonlinear1 commented 4 years ago

Thank you for your suggestion, for now, I am reading your codes. I want to remove ros and add imu. You tell me that you have tested against a few minute long ros-bag-files. The quality of depth picture is very good? I test OpenDTAM with a dataset made by me with orbslam2. I found the depth of picture is very bad. the pointcloud of OpenDTAM with my dataset is twisted!