mirror of
https://github.com/invoke-ai/InvokeAI
synced 2024-08-30 20:32:17 +00:00
55 lines
1.6 KiB
Python
55 lines
1.6 KiB
Python
import os
|
|
import torch
|
|
import cv2
|
|
import numpy as np
|
|
from PIL import Image
|
|
|
|
from diffusers.utils import load_image
|
|
from diffusers.models.controlnet import ControlNetModel
|
|
from invokeai.backend.generator import Txt2Img
|
|
from invokeai.backend.model_management import ModelManager
|
|
|
|
|
|
print("loading 'Girl with a Pearl Earring' image")
|
|
image = load_image(
|
|
"https://hf.co/datasets/huggingface/documentation-images/resolve/main/diffusers/input_image_vermeer.png"
|
|
)
|
|
image.show()
|
|
|
|
print("preprocessing image with Canny edge detection")
|
|
image_np = np.array(image)
|
|
low_threshold = 100
|
|
high_threshold = 200
|
|
canny_np = cv2.Canny(image_np, low_threshold, high_threshold)
|
|
canny_image = Image.fromarray(canny_np)
|
|
canny_image.show()
|
|
|
|
# using invokeai model management for base model
|
|
print("loading base model stable-diffusion-1.5")
|
|
model_config_path = os.getcwd() + "/../configs/models.yaml"
|
|
model_manager = ModelManager(model_config_path)
|
|
model = model_manager.get_model('stable-diffusion-1.5')
|
|
|
|
print("loading control model lllyasviel/sd-controlnet-canny")
|
|
canny_controlnet = ControlNetModel.from_pretrained("lllyasviel/sd-controlnet-canny",
|
|
torch_dtype=torch.float16).to("cuda")
|
|
|
|
print("testing Txt2Img() constructor with control_model arg")
|
|
txt2img_canny = Txt2Img(model, control_model=canny_controlnet)
|
|
|
|
print("testing Txt2Img.generate() with control_image arg")
|
|
outputs = txt2img_canny.generate(
|
|
prompt="old man",
|
|
control_image=canny_image,
|
|
control_weight=1.0,
|
|
seed=0,
|
|
num_steps=30,
|
|
precision="float16",
|
|
)
|
|
generate_output = next(outputs)
|
|
out_image = generate_output.image
|
|
out_image.show()
|
|
|
|
|
|
|