|
|
--- |
|
|
license: cc-by-4.0 |
|
|
language: |
|
|
- en |
|
|
base_model: |
|
|
- shuttleai/shuttle-3.1-aesthetic |
|
|
pipeline_tag: image-to-image |
|
|
--- |
|
|
# Fashion Edit Model |
|
|
|
|
|
This repository contains a specialized fashion image editing model built on top of Shuttle-3.1-Aesthetic. The model has been trained to excel at fashion-specific modifications, including style transfer, seasonal adaptations, and fabric/texture transformations. |
|
|
|
|
|
## Model Features |
|
|
|
|
|
- Built on Shuttle-3.1-Aesthetic architecture |
|
|
- Enhanced transformer with increased input channels (128 instead of 64) |
|
|
- Specialized for fashion-specific editing tasks |
|
|
|
|
|
## Using the Model |
|
|
|
|
|
This model works with our custom `FluxPix2PixPipeline` for optimal results: |
|
|
|
|
|
```python |
|
|
import torch |
|
|
from diffusers import DiffusionPipeline |
|
|
import PIL.Image |
|
|
|
|
|
# Load the model with custom pipeline |
|
|
pipe = DiffusionPipeline.from_pretrained( |
|
|
"aihpi/fashion-edit-model", |
|
|
custom_pipeline="aihpi/fashion-edit-model", |
|
|
torch_dtype=torch.bfloat16 |
|
|
) |
|
|
|
|
|
# Move to GPU if available |
|
|
device = "cuda" if torch.cuda.is_available() else "cpu" |
|
|
pipe = pipe.to(device) |
|
|
|
|
|
# Load an image |
|
|
input_image = PIL.Image.open("fashion_image.jpg").convert("RGB") |
|
|
input_image = input_image.resize((768, 768)) |
|
|
|
|
|
# Generate transformed image |
|
|
prompt = "Convert this outfit to winter style with wool textures" |
|
|
|
|
|
result = pipe( |
|
|
prompt=prompt, |
|
|
image=input_image, |
|
|
num_inference_steps=6, |
|
|
guidance_scale=7.5 |
|
|
).images[0] |
|
|
|
|
|
result.save("winter_outfit.png") |
|
|
``` |
|
|
|
|
|
## Model Card |
|
|
|
|
|
**Developer:** AIHPI |
|
|
|
|
|
**Base Model:** shuttleai/shuttle-3.1-aesthetic |
