---
license: apache-2.0
tags:
- travel
- africa
---
Predicts Africa landmarks from a set of 2214 classes. Output is logits.

Trained on data from: https://github.com/cvdfoundation/google-landmark

Classification head built on top of EfficentnetV2-M-21k feature extractor for [CrossPrism](https://apps.apple.com/us/app/crossprism-photo-labeler/id1638429352?mt=12) on MacOS.

Compressed using 8-bit quantization.