README.md

---
license: openrail
---


# ControlNet-XS

![gif](./teaser.gif)

These are ControlNet-XS weights trained on [stabilityai/stable-diffusion-xl-base-1.0](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0) and  [stabilityai/stable-diffusion-2-1](https://huggingface.co/stabilityai/stable-diffusion-2-1) on edge and depthmap conditioning respectively. You can find more details and further visual examples on the project page [ControlNet-XS](https://vislearn.github.io/ControlNet-XS/).

## The codebase
The code is based on on the StableDiffusion frameworks. To use the ControlNet-XS, you need to access the weights for the StableDiffusion version that you want to control separately.
We provide the weights with both depth and edge control for StableDiffusion2.1 and StableDiffusion-XL.

After obtaining the weights, you need the replace the paths to the weights of StableDiffusion and ControlNet-XS in the config files.
![images_1)](./banner_image.png)
## Usage


Example for StableDiffusion-XL with Canny Edges

```python
import scripts.control_utils as cu
import matplotlib.pyplot as plt
import torch
from PIL import Image

path_to_config = 'ControlNet-XS-main/configs/inference/sdxl/sdxl_encD_canny_48m.yaml'
model = cu.create_model(config_path_depth)

image_path = 'PATH/TO/IMAGES/Shoe.png'

canny_high_th = 250
canny_low_th = 100
size = 768
num_samples=2

image = cu.get_image(image_path, size=size)
edges = cu.get_canny_edges(image, low_th=canny_low_th, high_th=canny_high_th)

samples, controls = cu.get_sdxl_sample(
    guidance=edges,
    ddim_steps=10,
    num_samples=2,
    model=model,
    shape=[4, size // 8, size // 8],
    control_scale=0.95,
    prompt='cinematic, shoe in the streets, made from meat, photorealistic shoe, highly detailed',
    n_prompt='lowres, bad anatomy, worst quality, low quality',
)


Image.fromarray(cu.create_image_grid(samples)).save('SDXL_MyShoe.png')
```
![images_1)](./SDXL_MyShoe.png)

Example for StableDiffusion2.1 with depth maps


```python
import scripts.control_utils as cu
import matplotlib.pyplot as plt
import torch
from PIL import Image

path_to_config = 'PATH/TO/CONFIG/sd21_encD_depth_14m.yaml'
model = cu.create_model(path_to_config)

size = 768
image_path = 'PATH/TO/IMAGES/Shoe.png'


image = cu.get_image(image_path, size=size)
depth = cu.get_midas_depth(image, max_resolution=size)
num_samples = 2

samples, controls = cu.get_sd_sample(
    guidance=depth,
    ddim_steps=10,
    num_samples=num_samples,
    model=model,
    shape=[4, size // 8, size // 8],
    control_scale=0.95,
    prompt='cinematic, advertising shot, shoe in a city street, photorealistic shoe, colourful, highly detailed',
    n_prompt='low quality, bad quality, sketches'
)


Image.fromarray(cu.create_image_grid(samples)).save('SD_MyShoe.png')
```
![images_2)](./SD_MyShoe.png)