Marketplace

Small Vehicle Detection

Detects small vehicles on satellite images with 0.5m GSD.

Description

Small Vehicle Detection identifies small vehicles such as cars, vans, etc. in satellite images in both urban and rural settings. The block can detect vehicles in images with ground sampling distance (GSD) around 0.5 m. Output is provided as Geojson file containing detection results in Point feature format. The algorithm is trained on a data-set obtained from South-Eastern Asia.

Technical description

The algorithm uses deep learning techniques and Faster R-CNN detection architecture to detect small vehicles such as cars, 3-wheelers and SUVs.

Use cases

Traffic Monitoring, Transportation Infrastructure Design, Smart-City Application, Measuring Retail Traffic, Economic Activity, Hospital Traffic

Note: This block can be used with the Pléiades Streaming block, without the use of Raster Tiling. Pléiades Download is currently not supported.

General InformationDescription
Block TypeProcessing
Supported input dataInput is required as a set of Tagged Image File

Format (TIFF) image and the GeoJSON file. The image is expected to have Ground Sampling Distance (GSD) less than 0.55 m.| | Output data format | The output is point feature GeoJSON which contains points on detected vehicles.| | Algorithm Training Data Details | The algorithm is trained on a dataset obtained from South Asia. Most of the settings are dense urban areas, neighborhoods, parking lots, industrial parks, etc. | | Algorithm Performance | 0.5 IoU and has detection of 0.5mAP on satellite images with GSD 0.55m |

Fore more information, visit the provider website here.

Capabilities

Input Capabilities

raster
up42_standard
format
{
  "or": [
    "GTiff",
    "PNG",
    "JPEG"
  ]
}
resolution
0.5

Output Capabilities

vector
custom
object_type
small_vehicle
up42_standard
format
GeoJSON
geometry_type
Point

End User License Agreement

View the End User License Agreement conditions in a new tab.