Jordiett/convnextv2-geoguessr

ConvNeXt V2 Base Fine-tuned on GeoGuessr

Model Description

This model is a ConvNeXt V2 Base fine-tuned for geographic image classification of Google Street View images. The model can identify the country of origin of an image among 55 different countries.

ConvNeXt V2 is a modern evolution of convolutional networks that incorporates improvements inspired by Vision Transformers, while maintaining the efficiency and spatial inductive bias of traditional CNNs.

Results

The model has been trained and evaluated on the GeoGuessr dataset with the following results:

Metric	Value
Accuracy	61.03%
F1-Macro	51.77%

Comparison with Previous Work

Model	Accuracy	F1-Macro
ConvNeXt V2 (this model)	61.03%	51.77%
SigLIP2 (prithivMLmods)	64.85%	38.36%
ViT-Base-384 (dataset author)	38.81%	14.40%

Although SigLIP2 has slightly higher accuracy, this model achieves a significantly better F1-Macro (+13.41%), indicating more balanced performance across all classes.

Performance by Country

Top 10 best performing countries:

• 🇯🇵 Japan: 85.98%
• 🇧🇼 Botswana: 82.61%
• 🇩🇪 Germany: 80.18%
• 🇦🇺 Australia: 78.93%
• 🇿🇦 South Africa: 78.74%
• 🇨🇦 Canada: 76.26%
• 🇬🇧 United Kingdom: 74.06%
• 🇸🇬 Singapore: 72.63%
• 🇹🇭 Thailand: 70.34%
• 🇮🇱 Israel: 69.49%

Top 10 most challenging countries:

• 🇱🇻 Latvia: 14.29%
• 🇸🇰 Slovakia: 18.87%
• 🇵🇹 Portugal: 21.28%
• 🇱🇹 Lithuania: 21.43%
• 🇨🇿 Czechia: 22.38%
• 🇺🇦 Ukraine: 23.33%
• 🇭🇷 Croatia: 25.00%
• 🇩🇰 Denmark: 28.57%
• 🇧🇪 Belgium: 29.47%
• 🇮🇹 Italy: 32.23%