Algorithmia vegetation detection

Part of trees/ungp

This algorithm accepts as input a directory of images hosted on algorithmia, amazon s3 or dropbox and returns the percentage vegetation detected in each image.

Running

With algorithmia algo client:

algo run nocturne/vegetation/$(git rev-parse HEAD) -d \
  '{"src": "data://.my/test_images", "method": "deep"}'

The algo. accepts 2 arguments:

• src: The location of your street-level images.
• method: The segmentation method to use.

Segmentaton methods

There are 2 options to choose from.

• lab: This method will determine the percentage vegetation present in each image by counting the ratio of detected green pixels.
• deep: This method will make use of a deep image segmentation algorithm to derive percentage vegetation present in each image. (This is the preferred option).

Please see our paper for more details.

Example

# lab based segmentation
[phil@arasaka vegetation]$ algo run nocturne/vegetation/$(git rev-parse HEAD) -d \
	'{"src": "data://.my/test_images", "method": "lab"}'
Completed in 0.8 seconds
[0.0063,0.1172,0.0259]

# deep segmentation
[phil@arasaka vegetation]$ algo run nocturne/vegetation/$(git rev-parse HEAD) -d \
	'{"src": "data://.my/test_images", "method": "deep"}'
Completed in 8.7 seconds
[0.0087,0,0.0297]

Maintainer

• Phil Stubbings @DataSciCampus.