StreetCLIP is a robust foundation model for open-domain image geolocalization and other geographic and climate-related tasks.
Experience the model with the demo screen.
StreetCLIP deeply understands the visual features found in street-level urban and rural scenes and knows how to relate these concepts to specific countries, regions, and cities. Given its training setup, the following use cases are recommended for StreetCLIP.
StreetCLIP can be used out-of-the box using zero-shot learning to infer the geolocation of images on a country, region, or city level. Given that StreetCLIP was pretrained on a dataset of street-level urban and rural images, the best performance can be expected on images from a similar distribution.
Broader direct use cases are any zero-shot image classification tasks that rely on urban and rural street-level understanding or geographical information relating visual clues to their region of origin.
Understanding the Built Environment
- Analyzing building quality
- Building type classifcation
- Building energy efficiency Classification
- Analyzing road quality
- Utility pole maintenance
- Identifying damage from natural disasters or armed conflicts
Understanding the Natural Environment
- Mapping vegetation
- Vegetation classification
- Soil type classifcation
- Tracking deforestation
General Use Cases
- Street-level image segmentation
- Urban and rural scene classification
- Object detection in urban or rural environments
- Improving navigation and self-driving car technology