Semantic mapping based on pixel level classification.
The weight file should be placed in deeplabv2/data/models
,and downloads in deeplab-pytorch . And you can change the path in deeplabv2/inference.py
.
class SegModel():
def __init__(self):
config_path = "deeplabv2/configs/cocostuff164k.yaml"
model_path = "deeplabv2/data/models/deeplabv2_resnet101_msc-cocostuff164k-100000.pth"
......
The same as ORB-SLAM2.
Refer to 基于ORB-SLAM2的语义地图构建.