Importing segmentation data from VGG Image Annotator (VIA)

Nickelberry · September 23, 2019, 8:28am

I have a data set containing 10,000 segmented satellite images which were created in VGG Image Annotator, I would like to use these images for image segmentation as in the CamVid example in part 1 lesson 3. However I’m unsure of the best way to pull in the annotations. Could somebody please point me in the right direction?

This is the tool used for the annotation
http://www.robots.ox.ac.uk/~vgg/software/via/via.html

Nickelberry · September 26, 2019, 3:51am

Ok so I managed to sort this myself, if anyone in the future has this same problem my solution is below,
It’s not very nice code but it works, and is pretty quick. It only takes a couple seconds to process 10,000 annotations

gist.github.com

https://gist.github.com/wrignj08/06e69f6068a25967127b729eb03f8fcc

VGG VIA to Raster Mask.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Befor you start you need to set up a couple things.\n",
    "Make a new folder, within that folder you need 3 things:\n",
    "1 All of your images in a folder called \"images\"\n",
    "2 Your VGG VIA anotations exported as a csv\n",

This file has been truncated. show original