zhang373 / depth-estimation-based-on-pre-trained-models-for-slam-robot Goto Github PK

Dept. of Electrical and Electronic Engineering, University of Hong Kong, Prof. Even Peng

• Task: Use one camera to build an end to end system to better do depth estimation
• Use simple U-Net to do first experiment
• Use large pretrained feature extinction model to replace encoder part of U-Net and pretrained large up sampling model to enhance decoder part of U-Net. Do experiment on SLAM robort.