---
library_name: pytorch
license: mit
pipeline_tag: image-classification
tags:
- vision-transformer
- walkability
---

# ViT-Walkability

This repository contains a fine-tuned Vision Transformer model for walkability-related image classification tasks.

## Checkpoint details
- Source file: `best.pt`
- Top-level type: `dict`
- Base model: `google/vit-base-patch16-224`
- Num labels: `5`
- Labels: `['LABEL_0', 'LABEL_1', 'LABEL_2', 'LABEL_3', 'LABEL_4']`
- Keys: `['model_state_dict', 'epoch']`

## Loading notes
The uploaded files are in standard Hugging Face Transformers format and can be loaded directly with `from_pretrained`.