---
license: mit
tags:
- image-to-image
datasets:
- yulu2/InstructCV-Demo-Data
---

# InstructCV: Instruction-Tuned Text-to-Image Diffusion Models as Vision Generalists

GitHub: https://github.com/AlaaLab/InstructCV

[![pCVB5B8.png](https://s1.ax1x.com/2023/06/11/pCVB5B8.png)](https://imgse.com/i/pCVB5B8)


## Example

To use `InstructCV`, install `diffusers` using `main` for now. The pipeline will be available in the next release

```bash
pip install diffusers accelerate safetensors transformers
```

```python
import PIL
import requests
import torch
from diffusers import StableDiffusionInstructPix2PixPipeline, EulerAncestralDiscreteScheduler

model_id = "yulu2/InstructCV"
pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(model_id, torch_dtype=torch.float16, safety_checker=None, variant="ema")
pipe.to("cuda")
pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)

url = "put your url here"

def download_image(url):
    image = PIL.Image.open(requests.get(url, stream=True).raw)
    image = PIL.ImageOps.exif_transpose(image)
    image = image.convert("RGB")
    return image

image         = download_image(URL)
seed          = random.randint(0, 100000)
generator     = torch.manual_seed(seed)
width, height = image.size
factor        = 512 / max(width, height)
factor        = math.ceil(min(width, height) * factor / 64) * 64 / min(width, height)
width         = int((width * factor) // 64) * 64
height        = int((height * factor) // 64) * 64
image         = ImageOps.fit(image, (width, height), method=Image.Resampling.LANCZOS)

prompt        = "Detect the person."
images        = pipe(prompt, image=image, num_inference_steps=100, generator=generator).images[0]
images[0]
```