Reality Capture

Overview

The Reality Capture API lets you generate 3D models, point clouds, and orthophotos from photographs using photogrammetry. You provide a set of overlapping images of an object or scene, and APS processes them into a 3D model in your chosen output format.

Typical workflow:

1. Authenticate — get a 2-legged token.
2. Create a photoscene — define the output format.
3. Upload images — send your photographs to the photoscene.
4. Process — trigger photogrammetry processing.
5. Poll for completion — wait until the model is ready.
6. Download — retrieve the finished 3D model.

Authentication

library(AutoDeskR)

tok <- getToken(
  id     = Sys.getenv("client_id"),
  secret = Sys.getenv("client_secret"),
  scope  = "data:read data:write data:create"
)

Create a Photoscene

ps <- createPhotoscene(
  name   = "building-survey",
  format = "obj",
  token  = tok
)

photoscene_id <- ps$content$photoscene$photosceneid

Supported output formats:

Format	Description
rcm	ReCap point cloud (default)
rcs	ReCap scan
obj	Wavefront OBJ mesh
ortho	Orthophoto
report	Quality report only

Upload Images

Provide a character vector of local file paths. The Reality Capture API supports JPEG and PNG images. For best results, use 20–300 overlapping photos with at least 60% overlap between adjacent shots.

image_files <- list.files("~/survey-photos", pattern = "\\.jpg$", full.names = TRUE)

imgs <- uploadImages(
  photoscene_id = photoscene_id,
  files         = image_files,
  token         = tok
)

Start Processing

proc <- processPhotoscene(
  photoscene_id = photoscene_id,
  token         = tok
)

Wait for Completion

Processing is asynchronous and can take several minutes to hours depending on the number of images. Use waitForPhotoscene() to block until it finishes:

done <- waitForPhotoscene(
  photoscene_id = photoscene_id,
  token         = tok,
  interval      = 30,    # poll every 30 seconds
  timeout       = 3600,  # give up after 1 hour
  verbose       = TRUE
)

Or poll manually with checkPhotoscene():

status <- checkPhotoscene(photoscene_id = photoscene_id, token = tok)
status$content$photoscene$progress      # e.g. "75"
status$content$photoscene$progressmsg   # e.g. "Processing..."

Download the Result

Once processing reaches 100%, download the output via the Model Derivative API. First get the output URN from the manifest, then download the file:

encoded_urn    <- jsonlite::base64_enc(photoscene_id)
manifest       <- getOutputUrn(urn = encoded_urn, token = tok)

encoded_output <- jsonlite::base64_enc(manifest$content$derivatives[[1]]$children[[1]]$urn)
result <- downloadFile(
  urn        = encoded_urn,
  output_urn = encoded_output,
  token      = tok,
  destfile   = "building-survey.obj"
)

Tips

• Use images with consistent exposure and focus.
• Aerial (drone) images should have at least 80% overlap.
• The API supports up to 1,000 aerial or 300 hand-held images per photoscene.
• Processing time depends on the number of images and server load.