GeoCLIP Documentation

You can install the GeoCLIP module using pip or by installing directly from the source code.

# Via pip
pip install geoclip

# Or from source
git clone https://github.com/VicenteVivan/geo-clip
cd geo-clip
python setup.py install

Use the GeoCLIP class to predict GPS coordinates from an image. The predict method returns the top $k$ predicted GPS coordinates (latitude and longitude) and their corresponding probabilities.

import torch
from geoclip import GeoCLIP

model = GeoCLIP()

image_path = "image.png"

# Returns top k GPS predictions and probabilities
top_pred_gps, top_pred_prob = model.predict(image_path, top_k=5)

print("Top 5 GPS Predictions")
print("=====================")
for i in range(5):
    lat, lon = top_pred_gps[i]
    print(f"Prediction {i+1}: ({lat:.6f}, {lon:.6f})")
    print(f"Probability: {top_pred_prob[i]:.6f}")
    print("")

The LocationEncoder can be used to transform GPS coordinates (latitude and longitude) into semantically rich embeddings. These embeddings can be used to assist geo-aware neural architectures, such as concatenating them with visual features for improved multi-class classification.

import torch
from geoclip import LocationEncoder

gps_encoder = LocationEncoder()

# Input data as a tensor of [latitude, longitude]
gps_data = torch.Tensor([[40.7128, -74.0060], [34.0522, -118.2437]])  # NYC and LA

# Returns embeddings of shape (N, 512)
gps_embeddings = gps_encoder(gps_data)
print(gps_embeddings.shape) # (2, 512)