https://github.com/shunk031/simple-aesthetics-predictor
CLIP-based aesthetics predictor inspired by the interface of 🤗 huggingface transformers.
https://github.com/shunk031/simple-aesthetics-predictor
aesthetic-predictor clip huggingface huggingface-transformers vit
Last synced: about 1 year ago
JSON representation
CLIP-based aesthetics predictor inspired by the interface of 🤗 huggingface transformers.
- Host: GitHub
- URL: https://github.com/shunk031/simple-aesthetics-predictor
- Owner: shunk031
- License: mit
- Created: 2023-07-04T09:29:42.000Z (almost 3 years ago)
- Default Branch: master
- Last Pushed: 2024-06-14T11:20:24.000Z (almost 2 years ago)
- Last Synced: 2025-04-22T03:29:34.610Z (about 1 year ago)
- Topics: aesthetic-predictor, clip, huggingface, huggingface-transformers, vit
- Language: Python
- Homepage:
- Size: 532 KB
- Stars: 36
- Watchers: 2
- Forks: 0
- Open Issues: 4
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# 🤗 Simple Aesthetics Predictor
[](https://github.com/shunk031/simple-aesthetics-predictor/actions/workflows/ci.yaml)
[](https://github.com/shunk031/simple-aesthetics-predictor/actions/workflows/release.yaml)
[](https://github.com/shunk031/simple-aesthetics-predictor/actions/workflows/deploy.yaml)
[](https://pypi.python.org/pypi/simple-aesthetics-predictor)
[CLIP](https://arxiv.org/abs/2103.00020)-based aesthetics predictor inspired by the interface of [🤗 huggingface transformers](https://huggingface.co/docs/transformers/index).
This library provides a simple wrapper that can load the predictor using the `from_pretrained` method.
We currently provide the following wrappers for aesthetics predictor:
- **v1**: LAION-AI/aesthetic-predictor: A linear estimator on top of clip to predict the aesthetic quality of pictures https://github.com/LAION-AI/aesthetic-predictor
- **v2**: christophschuhmann/improved-aesthetic-predictor: CLIP+MLP Aesthetic Score Predictor https://github.com/christophschuhmann/improved-aesthetic-predictor
## Install
```shell
pip install simple-aesthetics-predictor
```
## How to Use
```python
import requests
import torch
from PIL import Image
from transformers import CLIPProcessor
from aesthetics_predictor import AestheticsPredictorV1
#
# Load the aesthetics predictor
#
model_id = "shunk031/aesthetics-predictor-v1-vit-large-patch14"
predictor = AestheticsPredictorV1.from_pretrained(model_id)
processor = CLIPProcessor.from_pretrained(model_id)
#
# Download sample image
#
url = "https://github.com/shunk031/simple-aesthetics-predictor/blob/master/assets/a-photo-of-an-astronaut-riding-a-horse.png?raw=true"
image = Image.open(requests.get(url, stream=True).raw)
#
# Preprocess the image
#
inputs = processor(images=image, return_tensors="pt")
#
# Move to GPU
#
device = "cuda"
predictor = predictor.to(device)
inputs = {k: v.to(device) for k, v in inputs.items()}
#
# Inference for the image
#
with torch.no_grad(): # or `torch.inference_model` in torch 1.9+
outputs = predictor(**inputs)
prediction = outputs.logits
print(f"Aesthetics score: {prediction}")
```
## The Predictors found in 🤗 Huggingface Hub
- 🤗 [aesthetics-predictor-v1](https://huggingface.co/models?search=aesthetics-predictor-v1)
- 🤗 [aesthetics-predictor-v2](https://huggingface.co/models?search=aesthetics-predictor-v2)
## Acknowledgements
- LAION-AI/aesthetic-predictor: A linear estimator on top of clip to predict the aesthetic quality of pictures https://github.com/LAION-AI/aesthetic-predictor
- christophschuhmann/improved-aesthetic-predictor: CLIP+MLP Aesthetic Score Predictor https://github.com/christophschuhmann/improved-aesthetic-predictor