test

app.py

@@ -1,4 +1,10 @@
 import gradio as gr
+from diffusers import StableDiffusionPipeline, ControlNetModel, StableDiffusionControlNetPipeline
+from diffusers.utils import load_image
+import torch
+import cv2
+import numpy as np
+from PIL import Image
 
 is_show_controlnet = True
 prompts = ""
@@ -10,8 +16,38 @@ def change_radio(input):
 def output_radio(output):
     print(output)
 
+def predict(canny, lt, ht, prompt, neg_prompt, ins, gs, seed):
+    print(canny, lt, ht, prompt, neg_prompt, ins, gs)
+    '''
+    np_image = np.array(canny)
+
+    low_threshold = lt
+    high_threshold = ht
+
+    np_image = cv2.Canny(np_image, low_threshold, high_threshold)
+    np_image = np_image[:, :, None]
+    np_image = np.concatenate([np_image, np_image, np_image], axis=2)
+    canny_image = Image.fromarray(np_image)
+    controlnet_repo_id = "calihyper/kor-portrait-controlnet"
+    controlnet = ControlNetModel.from_pretrained(controlnet_repo_id, torch_dtype=torch.float16)
+    '''
+    repo_id = "calihyper/trad-kor-landscape-black"
+    pipe = StableDiffusionControlNetPipeline.from_pretrained(
+        repo_id, torch_dtype=torch.float16
+    )
+    generator = torch.manual_seed(seed)
+
+    output = pipe(
+        prompt,
+        negative_prompt=neg_prompt,
+        generator=generator,
+        num_inference_steps=ins,
+        guidance_scale=gs
+    )
+    return output.images[0]
+
 with gr.Blocks() as demo:
-    gr.Markdown("# Greetings from Gradio!")
+    gr.Markdown("# Aiffelthon Choosa Project")
 
     with gr.Row():
         with gr.Column() as controlnet:
@@ -19,19 +55,21 @@ with gr.Blocks() as demo:
             canny_image = gr.Image(label="cannyimage", visible=is_show_controlnet , shape=(512,512), interactive=True)
         
             controlnet_radio = gr.Radio([True, False], label="Use ControlNet")
-            lt = gr.Slider(1, 50, 3, step=1, label="Low threshold")
-            ht = gr.Slider(1, 50, 3, step=1, label="High threshold")        
+            lt = gr.Slider(50, 300, 120, step=1, label="Low threshold")
+            ht = gr.Slider(50, 300, 120, step=1, label="High threshold")        
 
         with gr.Column():
             out_image = gr.Image()
+            with gr.Column() as diff:
+                prompt = gr.Textbox(placeholder="prompts", label="prompt")
+                neg_prompt = gr.Textbox(placeholder="negative prompts", value=neg_prompt, label="negative prompt")
 
-            prompt = gr.Textbox(placeholder="prompts", label="prompt")
-            neg_prompt = gr.Textbox(placeholder="negative prompts", label="negative prompt")
-
-            ins = gr.Slider(1, 50, 3, label="inference steps")
-            gs = gr.Slider(1, 50, 3, label="guidance scale")
+                ins = gr.Slider(1, 60, 30, label="inference steps")
+                gs = gr.Slider(1, 10, 2.5, step=1, label="guidance scale")
 
+                seed = gr.Slider(0, 10, 2, step=1, label="seed")
     btn1 = gr.Button("실행")
+    btn1.click(predict, [canny_image, lt, ht, prompt, neg_prompt, ins, gs, seed], out_image)
 
 if __name__ == "__main__":
     demo.launch()