Home

Awesome

Active Vision Dataset Processing

If you are looking for the old repository, pre March 2018, it is here

This project contains Python and MATLAB scripts for processing the Active Vision Dataset

DATA LOADING

A starting point for loading our data in python for detection or classification of instances. See data_loading/demo.py

With this code, you can:

  1. Get an iterable data structure that gives images and bounding box labels
  2. Use the data structure to crop boxes for classification
  3. Apply pre-defined and custom transforms to the images and labels

Visualizations

Use the code here to visualize our data. See the README in this folder for examples.

To get started, edit the ROHIT_BASE_PATH variable in run.py and/or init_paths.m to point to the directory that contains our data. i.e. the parent directory that the directory for each scan.

  ROHIT_BASE_PATH = /path/to/dataset/

Running visualization code

Everything is a function. See examples below for how to run certain functions, or look at the README in the visualizations directory

Python

To run a python function from the command line:

python run.py function_name scene_name

MATLAB

  1. Start MATLAB

  2. Run init_paths

  3. Call the desired function