@heethesh Hi, thanks for the inspiring work! I would like to know the right way to run your model on the TUM dataset( e.g. walking). The following is how I'm going to try it, and please correct me if I misunderstand:

1. catkin_make orb_slam(I use the one in your repo)
2. segment the pictures converted from the bag (I would appreciate if the steps are offered)
3. run updata_rosbag to apply the semantic segmentation result, since the script requiresimage-dir (I would like to know the specific way to do this step)

Thanks again for the great work, looking forward to your reply.

Hello.

THanks for your code. I was able to launch salsa on mono euroc data but i've got some strange results. Here

THing is that there are like no semantic labels on the pointcloud. Just white points. I'm launching eouroc bag, which i've updated with segmentation as it said in FAQ, using commands:
rosbag play V1_01_easy_segment.bag /cam0/image_raw:=/d400/color/image_raw /cam0/dynamic_scores:=/d400/annotation/dynamic_scores
roslaunch orb_slam2_ros salsa.launch (by the way. In your FAQ it is orb_slam_2_ros which is incorrect in my case).

ANy suggestions what i'm doing wrong to see semantically labelled pointcloud?