BertaBescos / DynaSLAM

DynaSLAM is a SLAM system robust in dynamic environments for monocular, stereo and RGB-D setups
https://bertabescos.github.io/DynaSLAM/
Other
829 stars 201 forks source link

Output matrix of object mask #76

Open Rushmian opened 3 years ago

Rushmian commented 3 years ago

Hi,

Thanks for this wonderful dynamic object based SLAM system. I wanted to know something regarding the object mask. So far I understand, after detection of dynamic objects using Mask-RCNN, an output of object mask matrix is obtained. If there is only one class of object in the frame, it will give a mask matrix of 0's and the class-ID for that particular class of object (e.g. for cars, the class_ID is 3).

If I wanted to get an output of the segmentation in a frame, like the image I have attached, I would find that the objects of same class will have a similar shade of color. But if I wanted to give different shades of color for different objects of same class, what can I do about it? Basically, if there are two cars in the frames, I wanted to get different values for the two cars in the output object mask matrix.

p.s. I am using a KITTI dataset which is given below -

https://drive.google.com/file/d/1LpjIdh6xL_UtWOkiJng0CKSmP7qAQhGu/view

semantic segmentation