Hello,
I am facing an issue where while running the mono or stereo node through ROS, I am getting a gray-scale image in tracking display instead of color.
I am sending BGR8 packed image format from my camera. When I display the image after using OpenCV, the images are colored. The camera used is Baumer VLG-20C.I
I know it must check for 3 channels but I am unable to understand why is it not detecting 3 channels.
Any help will be appreciated.
@lokeshluckybisht
ORB-SLAM2 uses FAST and BRIEF, which work on grayscale. Your color images are converted to grayscale as the very first step. Viewer shows orb-slam2 input image, not the one you supply.
@lokeshluckybisht
ORB-SLAM2 uses FAST and BRIEF, which work on grayscale. Your color images are converted to grayscale as the very first step. Viewer shows orb-slam2 input image, not the one you supply.
Thank you !!
Hello @AlejandroSilvestri , I wanted to know if depth information is being used while relocalization ?
I know that the code is somewhat different for RGBD and Mono upto creation of a frame, but afterwards, the code is same for both type of images. I wanted to know that If I have created and saved a map with RGBD camera, is it possible to relocalize with images taken from Mono RGB cam ? ( when I load the map back, it doesn't have any information wether the loaded map was created using a Mono, stereo or RGBD , then can I localize a RGB image ? )
@bhaveshachhada
Well, ORB-SLAM2 doesn't load maps, so it highly depends on what library are you using.
From my knowledge on monocular, I strongly believe the first part in relocalization works the same way: it uses the inverted database of BoW to quickly recover the most likely keyframes, those similar to current frame. Then there is a pose fitting that can be different. But the info required from the map must be the same in both cases, RGBD and monocular.
So, I believe you can localize a RGB image on a stereo or RGBD based map.
Thank You @AlejandroSilvestri Sir.