In MilesMcBain/r2vr: Compose and Serve WebVR Scenes

  knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.height = 3,
  fig.width = 7
  )
require(knitr)

Annotate 2D Virtual Reality Images by Creating an Interactive WebVR scene

r2vr has been extended to be able to help train users to annotate virtual reality images more accurately. The 2D Training will randomly select 3 images from a pool of images. Users are able to annotate selected markers (for each image) as either C or N to classify the bounded region as either containing mostly Coral or Not Coral respectively. The user's annotations are then compared to a gold standard, which are the correct annotations for the corresponding fixed points in which the user annotated. The user will then be able to receive visual feedback with markers changing to green or red depending on if they annotated the marker correctly or incorrectly respectively.

Case Study: Coral Cover

library(r2vr)
IPv4_ADDRESS <- find_IP() # Note: If not on Windows, enter IP directly

## TODO: SET full name here
# set_user("Firstname-Lastname") # default to be overridden
set_user("Jon-Peppinck")

## OPTIONAL: '?set_marker_and_props' shows configuration options
# i.e. Number of markers and size of markers, but keep "2d"
# e.g. set_marker_and_props("2d", 15, "small") 
set_marker_and_props("2d") 


## OPTIONAL: '?set_colors'
# e.g. set_colors(coral = "#FFFF00", not_coral = "#FF00FF", evaluation_selection = "#0000FF")
set_colors()

## TODO: SET the 'Gold Standard points' for the corresponding 'img_paths' (set below)
# Note: images are 4000x3000 (px) i.e. 0 <= x <= 4000, 0 <= y <= 3000
img1Points = list(
  list(id = 1, x = 3203, y = 173, isCoral = 0), ## sand?
  list(id = 2, x = 1726, y = 356, isCoral = 0),
  list(id = 3, x = 2291, y = 1086, isCoral = 0)
)

img2Points = list(
  list(id = 1, x = 1000, y = 1000, isCoral = 0),
  list(id = 2, x = 2000, y = 2000, isCoral = 0)
)

img3Points = list(
  list(id = 1, x = 300, y = 300, isCoral = 0),
  list(id = 2, x = 800, y = 800, isCoral = 0)
)

img4Points = list(
  list(id = 1, x = 800, y = 800, isCoral = 0),
  list(id = 2, x = 1500, y = 1500, isCoral = 0)
)

# NOTE: The center point of the marker will be at both corners => enough space needs to be allowed for the marker and potentially the menu options as well
img5Points = list(
  list(id = 1, x = 0, y = 0, isCoral = 0),
  list(id = 2, x = 4000, y = 3000, isCoral = 0)
)

R2VR_CDN <- "https://cdn.jsdelivr.net/gh/ACEMS/r2vr@experiment" # NOTE: Subject to change

R2VR_2D_IMAGES <- paste0(R2VR_CDN, "/inst/ext/images/2d/")

# TODO: Select images (4000x3000px)
# NOTE: If have other local images on PC can change img_paths to be a vector of relative file location for the current working directory
img_paths <- paste0(
  R2VR_2D_IMAGES,
  c("49001074001.jpeg",
    "49002256001.jpeg",
    "51010026001.jpeg",
    "49004035001.jpeg",
    "50003181001.jpeg")
)

img_paths_and_points <- list(
  # 2D image paths  4000x3000
  list(img = img_paths[1], img_points = img1Points),
  list(img = img_paths[2], img_points = img2Points),
  list(img = img_paths[3], img_points = img3Points),
  list(img = img_paths[4], img_points = img4Points),
  list(img = img_paths[5], img_points = img5Points)
)

set_random_images(img_paths_and_points)

animals <- shared_setup_scene("2d", "training") # DON'T CHANGE

COMMANDS - 2D TRAINING

start() fixed_markers() go_to() go_to() check(1) check(2) check(3) end() data.df <- read("https://r2vr.herokuapp.com/api/2d/training") # TODO: deploy rm(list=ls())

MilesMcBain/r2vr documentation built on March 29, 2021, 12:03 p.m.