---
license: other
license_name: stem.ai.mtl
license_link: LICENSE
tags:
- vision
- image-classification
- STEM-AI-mtl/City_map
- Google
- ViT
- STEM-AI-mtl
datasets:
- STEM-AI-mtl/City_map

---

# The fine-tuned ViT model that beats [Google's state-of-the-art model](https://huggingface.co/google/vit-base-patch16-224) and OpenAI's famous GPT4 for maps of cities around the world

Image-classification fine-tuned model that identifies which city map is illustrated from an image input.

The Vision Transformer (ViT) base model is a transformer encoder model (BERT-like) pretrained on a large collection of images in a supervised fashion, namely ImageNet-21k, at a resolution of 224x224 pixels. Next, the model was fine-tuned on ImageNet (also referred to as ILSVRC2012), a dataset comprising 1 million images and 1,000 classes, also at resolution 224x224.

- **Developed by:** STEM.AI
- **Model type:** Image classification of maps of cities
- **Finetuned from model:** [google/vit-base-patch16-224](https://huggingface.co/google/vit-base-patch16-224)


### How to use:

[Inference script](https://github.com/STEM-ai/Vision/blob/7d92c8daa388eb74e8c336f2d0d3942722fec3c6/ViT_inference.py)

For more code examples, we refer to [ViTdocumentation](https://huggingface.co/transformers/model_doc/vit.html#).

## Training data

This [Google's ViT-base-patch16-224 for city identification](https://huggingface.co/google/vit-base-patch16-224) model was fine-tuned on the [STEM-AI-mtl/City_map dataset](https://huggingface.co/datasets/STEM-AI-mtl/City_map), contaning overer 600 images of 45 different maps of cities around the world. 

## Training procedure

A Transformer training was performed on [google/vit-base-patch16-224](https://huggingface.co/google/vit-base-patch16-224) on a 4 Gb Nvidia GTX 1650 GPU.

[Training notebook](https://github.com/STEM-ai/Vision/raw/7d92c8daa388eb74e8c336f2d0d3942722fec3c6/Trainer_ViT.ipynb)

## Training evaluation results

The most accurate output model was obtained from a learning rate of 1e-3. The quality of the training was evaluated with the training dataset and resulted in the following metrics:

{'eval_loss': 1.3691096305847168,\
 'eval_accuracy': 0.6666666666666666,\
 'eval_runtime': 13.0277,\
 'eval_samples_per_second': 4.606,\
 'eval_steps_per_second': 0.154,\
 'epoch': 2.82}


## Model Card Authors

STEM.AI: stem.ai.mtl@gmail.com\
[William Harbec](https://www.linkedin.com/in/william-harbec-56a262248/)