add threshold and negative prompt

app.py

@@ -10,10 +10,10 @@ import cv2
 import torch
 
 from diffusers import StableDiffusion3ControlNetPipeline
-from diffusers.models import SD3ControlNetModel, SD3MultiControlNetModel
+from diffusers.models import SD3ControlNetModel
 from diffusers.utils import load_image
 
-# load pipeline
+# Load pipeline
 controlnet_canny = SD3ControlNetModel.from_pretrained("InstantX/SD3-Controlnet-Canny")
 pipe = StableDiffusion3ControlNetPipeline.from_pretrained(
     "stabilityai/stable-diffusion-3-medium-diffusers",
@@ -41,38 +41,35 @@ def resize_image(input_path, output_path, target_height):
 
 
 @spaces.GPU(duration=90)
-def infer(image_in, prompt, inference_steps, guidance_scale, control_weight, progress=gr.Progress(track_tqdm=True)):
-
-    n_prompt = 'NSFW, nude, naked, porn, ugly'
+def infer(image_in, prompt, negative_prompt, inference_steps, guidance_scale, control_weight, low_threshold, high_threshold, progress=gr.Progress(track_tqdm=True)):
 
     # Canny preprocessing
     image_to_canny = load_image(image_in)
     image_to_canny = np.array(image_to_canny)
-    image_to_canny = cv2.Canny(image_to_canny, 100, 200)
+    image_to_canny = cv2.Canny(image_to_canny, low_threshold, high_threshold)
     image_to_canny = image_to_canny[:, :, None]
     image_to_canny = np.concatenate([image_to_canny, image_to_canny, image_to_canny], axis=2)
     image_to_canny = Image.fromarray(image_to_canny)
 
     control_image = image_to_canny
  
-    # infer
+    # Infer
     image = pipe(
         prompt=prompt,
-        negative_prompt=n_prompt,
+        negative_prompt=negative_prompt,
         control_image=control_image, 
         controlnet_conditioning_scale=control_weight,
         num_inference_steps=inference_steps,
         guidance_scale=guidance_scale,
     ).images[0]
 
-    
     image_redim, w, h = resize_image(image_in, "resized_input.jpg", 1024)
     image = image.resize((w, h), Image.LANCZOS)
     
     return image, gr.update(value=image_to_canny, visible=True)
-   
 
-css="""
+
+css = """
 #col-container{
     margin: 0 auto;
     max-width: 1080px;
@@ -92,6 +89,7 @@ with gr.Blocks(css=css) as demo:
                 with gr.Column():
                     image_in = gr.Image(label="Image reference", sources=["upload"], type="filepath")
                     prompt = gr.Textbox(label="Prompt")
+                    negative_prompt = gr.Textbox(label="Negative Prompt", placeholder="Enter negative prompts here")
                     
                     with gr.Accordion("Advanced settings", open=False):
                         with gr.Column():
@@ -99,22 +97,23 @@ with gr.Blocks(css=css) as demo:
                                 inference_steps = gr.Slider(label="Inference steps", minimum=1, maximum=50, step=1, value=25)
                                 guidance_scale = gr.Slider(label="Guidance scale", minimum=1.0, maximum=10.0, step=0.1, value=7.0)
                             control_weight = gr.Slider(label="Control Weight", minimum=0.0, maximum=1.0, step=0.01, value=0.7)
+                            low_threshold = gr.Slider(label="Canny Low Threshold", minimum=0, maximum=255, step=1, value=100)
+                            high_threshold = gr.Slider(label="Canny High Threshold", minimum=0, maximum=255, step=1, value=200)
                     
                     submit_canny_btn = gr.Button("Submit")
                     
                 with gr.Column():
                     result = gr.Image(label="Result")
                     canny_used = gr.Image(label="Preprocessed Canny", visible=False)
-                    
 
 
     submit_canny_btn.click(
         fn=infer,
-        inputs=[image_in, prompt, inference_steps, guidance_scale, control_weight],
+        inputs=[image_in, prompt, negative_prompt, inference_steps, guidance_scale, control_weight, low_threshold, high_threshold],
         outputs=[result, canny_used],
         api_name="predict",
         show_api=True
     )
 
-# Enable API by setting enable_api=True
 demo.queue().launch(show_api=True)
+