BTS Pipeline

From Big to Small: Multi-Scale Local Planar Guidance for Monocular Depth Estimation
arXiv
Supplementary material

Video Demo

Purpose

This pipeline is used to test the performance of the BTS depth estimation model. It is subscribed to the camera/kitti topic and passes the incoming images to the depth model. The mock_publisher.py is responsible for publishing the images.

Start Depth Pipeline

Open a terminal and run:

$ roscore

On a new terminal, run the following code in your terminal once cd in the directory.

$ python2 main.py

Start Mock Publishing

On a new terminal, run the following code in your terminal once cd in the directory.

$ python2 mock_publisher.py config_test.txt

Where, the config_test.txt contains the path to the KITTI images to be published.

config_test.txt

The config file provides command line utilities in a file format for ease of use. The file contains:

--encoder densenet161_bts 
--data_path /home/mcav/DATA/kitti_dataset 
--image_path /2011_09_26/2011_09_26_drive_0022_sync/