base_model:
- black-forest-labs/FLUX.1-dev
library_name: diffusers
license_name: flux-1-dev-non-commercial-license
license_link: https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md
pipeline_tag: image-to-image
tags:
- ControlNet
- super-resolution
- upscaler
⚡ Flux.1-dev: Upscaler ControlNet ⚡
This is Flux.1-dev ControlNet for low resolution images developped by Jasper research team.
How to use
This model can be used directly with the diffusers library
import torch
from diffusers.utils import load_image
from diffusers import FluxControlNetModel
from diffusers.pipelines import FluxControlNetPipeline
# Load pipeline
controlnet = FluxControlNetModel.from_pretrained(
"jasperai/Flux.1-dev-Controlnet-Upscaler",
torch_dtype=torch.bfloat16
)
pipe = FluxControlNetPipeline.from_pretrained(
"black-forest-labs/FLUX.1-dev",
controlnet=controlnet,
torch_dtype=torch.bfloat16
)
# Load a control image
control_image = load_image(
"https://huggingface.co/jasperai/Flux.1-dev-Controlnet-Upscaler/resolve/main/examples/depth.jpg"
)
w, h = control_image.size
# Upscale x4
control_image = control_image.resize((w * 4, h * 4))
image = pipe(
"",
control_image=control_image,
controlnet_conditioning_scale=0.6,
num_inference_steps=28,
guidance_scale=3.5,
height=control_image.size[1],
width=control_image.size[0]
).images[0]
image
💡 Note: You can compute the conditioning map using for instance the MidasDetector from the controlnet_aux library
from controlnet_aux import MidasDetector
from diffusers.utils import load_image
midas = MidasDetector.from_pretrained("lllyasviel/Annotators")
# Load an image
im = load_image(
"https://huggingface.co/jasperai/jasperai/Flux.1-dev-Controlnet-Depth/resolve/main/examples/output.jpg"
)
surface = midas(im)
Training
This model was trained with a synthetic complex data degradation scheme taking as input a real-life image and artificially degrading it by combining several degradations such as amongst other image noising (Gaussian, Poisson), image blurring and JPEG compression. In a similar spirit as [1]
[1] Wang, Xintao, et al. "Real-esrgan: Training real-world blind super-resolution with pure synthetic data." Proceedings of the IEEE/CVF international conference on computer vision. 2021.
Licence
The licence under the Flux.1-dev model applies to this model.