Depth map estimation from monocular camera

Is anyone aware of ROS nodes implementing an algorithm to estimate a depth map from a monocular camera, like described in this paper?

I've found a little published code such as this and this and this, but it's all buggy Matlab code and undocumented.

Asked by Cerin on 2016-08-17 14:05:08 UTC

Comments

Answers

Hi Cerin ... I'm sorry don't have an answer for you but just wanted to connect to someone else trying to use Monocular Slam since I'm not getting any answers to my question posed several months ago [... search for this string in the Answers Forum: "How to display a 3D Map on RVIZ using ORB_SLAM2?". Please read that to understand where I am with ROS and Slam. Are you using ORB SLAM or ORB SLAM2? It's my understanding that you only get a depth 'scaled' map using a monocular camera since there is no Lidar or other range measurement device. The article links you provided are very interesting as an approach to extract depth from a singe camera. As a Software Developer, I've done some image processing in the past to extract image features for automatic object recognition and classification for the defense industry. If you think we can help each other, please contact me at pbarile@gmail.com. Thanks, Phil

Asked by pbarile on 2016-08-30 08:01:01 UTC

Comments

Hi. I haven't tested ORB SLAM, but I was able to get the PTAM ROS module working, which implements a similar algorithm. Performance was mixed, but I'm looking for methods of estimating depth directly in front of the path of movement, not tangentially, as these algorithms seem to do.

Asked by Cerin on 2016-08-30 18:31:13 UTC