search

kminoda / VIODE

VIODE: A Simulated Dataset to Address the Challenges of Visual-Inertial Odometry in Dynamic Environments
92 stars 9 forks source link

Regd. segmentation IDs #3

Closed karnikram closed 3 years ago

karnikram commented 3 years ago

The RGB to object IDs are provided in the rgb_ids.txt, but is it possible for you to provide the mapping between the different object classes and their object IDs?

kminoda commented 3 years ago

Thanks for the comment! I am sorry that the mapping for all the objects, but I have added the mapping between ID and vehicle classes.

wjf1997 commented 3 years ago

The RGB to object IDs are provided in the rgb_ids.txt, but is it possible for you to provide the mapping between the different object classes and their object IDs?

Hey bros, Have you ever processed semantic image? When I transformed the semantic image to single-channel, the value for each pixel cann't match the ID in the rgb_ids.txt. For example, the ID value of a vehicle from single-channel image is 109 but I cann't find 109 in other .txt files. (Actually all vehicles IDs is more than 240?) I am a rookie and looking forward to your answer.