ml-lambda-tiler

AWS Lambda + rio-tiler + STAC to provide Machine Learning Results to HOTOSM Tasking Manager

Info

---

Installation

Requirement

• AWS Account
• Docker
• node + npm

Create the package

# Build Amazon linux AMI docker container + Install Python modules + create package
git clone https://github.com/SpaceNetChallenge/ml-lambda-api.git
cd ml-lambda-all
make all

Deploy to AWS

#configure serverless (https://serverless.com/framework/docs/providers/aws/guide/credentials/)
# Create account with permissions for SLS deployment
# Record api endpoint
npm install
sls deploy

Endpoint

STAC API

/stac/bounds

Inputs:

• url: any valid url

example:

$ curl {you-endpoint}/stac/bounds?url=https://any-stac-item.json

  {"url": "https://any-file.on/the-internet.tif", "bounds": [90.47546096087822, 23.803014490532913, 90.48441996322644, 23.80577697976369]}

/stac/tiles/z/x/y.png

Inputs:

• url: any valid url
• asset_key: asset key from STAC_ITEM : default = raster Options:
• rgb: select bands indexes to return (e.g: (1,2,3), (4,1,2))
• nodata: nodata value to create mask

example:

$ curl {you-endpoint}/stac/tiles/7/10/10.png?url=https://any-stac-item.json

/stac/summary/z/x/y.json

Inputs:

• url: any valid url
• asset_key: asset key from STAC_ITEM : default = segmentation_mask

Outputs:

• pixel_count: Number of pixels of value out of 255
• object_count: Number of objects after polygonization

example:

$ curl {you-endpoint}/stac/tiles/7/10/10.png?url=https://any-stac-item.json&asset_key=segmentation_mask
  { 
  "pixel_count": 555,
  "object_count": 1000,
  }

COG API

/bounds

Inputs:

• url: any valid url

example:

$ curl {you-endpoint}/bounds?url=https://any-file.on/the-internet.tif

  {"url": "https://any-file.on/the-internet.tif", "bounds": [90.47546096087822, 23.803014490532913, 90.48441996322644, 23.80577697976369]}

/tiles/z/x/y.png

Inputs:

• url: any valid url

Options:

• rgb: select bands indexes to return (e.g: (1,2,3), (4,1,2))
• nodata: nodata value to create mask

example:

$ curl {you-endpoint}/tiles/7/10/10.png?url=https://any-file.on/the-internet.tif