Added depthmap tab

app.py

@@ -654,8 +654,167 @@ with gr.Blocks() as canny_blocks_interface:
 
 
 
+# ----- Depth Map Tab -----------------------------------------------------------------
+
+from diffusers import StableDiffusionControlNetPipeline, ControlNetModel, UniPCMultistepScheduler
+from controlnet_aux import CannyDetector, ContentShuffleDetector, HEDdetector, LineartAnimeDetector, LineartDetector, MidasDetector, MLSDdetector, NormalBaeDetector, OpenposeDetector, PidiNetDetector
+from PIL import Image, ImageChops, ImageOps
+from diffusers.utils import load_image
+from transformers import pipeline
+import numpy as np
+import requests
+import torch
+import cv2
+    
+def resize_image(image, max_dimension, multiplier=16):
+    original_width, original_height = image.size
+    aspect_ratio = original_width / original_height
+
+    if original_width > original_height:
+        new_width = min(max_dimension, original_width)
+        new_height = round(new_width / aspect_ratio)
+    else:
+        new_height = min(max_dimension, original_height)
+        new_width = round(new_height * aspect_ratio)
+
+    new_width = round(new_width / multiplier) * multiplier
+    new_height = round(new_height / multiplier) * multiplier
+    resized_image = image.resize((new_width, new_height), Image.ANTIALIAS)
+
+    return resized_image
+
+def depth_map_prompt(prompt, image_url, controlnet_pipe, controlnet_model, negative_prompt):
+  image = load_image(image_url)
+
+  max_dimension = 768
+  resized_image = resize_image(image, max_dimension)
+
+  depth_map = controlnet_model(resized_image)
+
+  output = controlnet_pipe(
+      prompt,
+      depth_map,
+      negative_prompt=negative_prompt,
+      generator=torch.Generator(device="cpu").manual_seed(2),
+      num_inference_steps=20,
+  )
+
+  return {"output": output.images[0], "depth_map": depth_map}
+
+
+
+
+controlnet_depth = ControlNetModel.from_pretrained(
+    "fusing/stable-diffusion-v1-5-controlnet-depth", torch_dtype=torch.float16
+)
+
+model_id = "runwayml/stable-diffusion-v1-5"
+depth_pipe = StableDiffusionControlNetPipeline.from_pretrained(
+    model_id,
+    controlnet=controlnet_depth,
+    torch_dtype=torch.float16,
+)
+
+depth_pipe.scheduler = UniPCMultistepScheduler.from_config(depth_pipe.scheduler.config)
+depth_pipe.enable_model_cpu_offload()
+depth_pipe.enable_xformers_memory_efficient_attention()
+
+loaded_model = MidasDetector.from_pretrained("lllyasviel/ControlNet") # works
+
+
+
+
+def rotate_image(image, rotation):
+  rotation = 360 - int(rotation)
+  image = image.rotate(rotation, resample=Image.BICUBIC, expand=True)
+  return image
+
+def controlnet_function(input_prompt, input_image, input_negative_prompt, input_seed, input_rotate, input_invert):
+  pil_image = Image.fromarray(input_image)
+
+  max_dimension = 768
+  processed_image = resize_image(pil_image, max_dimension, 32)
+
+  # rotate image
+  if input_rotate and int(input_rotate) > 0:
+    processed_image = rotate_image(processed_image, int(input_rotate))
+
+  depth_map = loaded_model(processed_image)
+
+  if input_invert:
+    depth_map = np.array(depth_map)
+    depth_map = 255 - depth_map
+    depth_map = Image.fromarray(depth_map)
+
+  generator = torch.Generator(device="cpu").manual_seed(input_seed)
+
+  output = depth_pipe(
+      input_prompt,
+      depth_map,
+      negative_prompt=input_negative_prompt,
+      generator=generator,
+      num_inference_steps=20,
+  )
+
+  return_text = f'''
+    prompt: "{input_prompt}"
+    seed: {input_seed}
+    negative-prompt: "{input_negative_prompt}"
+    controlnet: "fusing/stable-diffusion-v1-5-controlnet-depth"
+    stable-diffusion: "runwayml/stable-diffusion-v1-5"
+    inverted: {input_invert}
+  '''
+
+  return [return_text, output.images[0], depth_map]
+
+# import random 
+def random_seed():
+  return random.randint(0, 99999999999999)
+
+with gr.Blocks() as depth_controlnet_gradio:
+  gr.Markdown('''
+    # <span style="display: inline-block; height: 30px; width: 30px; margin-bottom: -3px; border-radius: 7px; background-size: 50px; background-position: center; background-image: url(http://www.astronaut.horse/thumbnail.jpg)"></span> ControlNet + Depthmap 
+    ---
+  ''')
+  with gr.Row():
+    with gr.Column():
+        gr.Markdown('''
+          ## Inputs...
+        ''')
+        input_prompt = gr.inputs.Textbox(label="text prompt")
+        input_image = gr.inputs.Image(label="input image")
+        with gr.Accordion(label="options", open=False):
+          with gr.Row():
+            with gr.Column():
+              input_negative_prompt = gr.inputs.Textbox(label="negative prompt")
+            with gr.Column():
+              input_seed = gr.Slider(0, 99999999999999, label="seed", dtype=int, value=random_seed, interactive=True, step=1)
+          with gr.Row():
+            with gr.Column():
+              input_rotate = gr.Dropdown([0, 90, 180, 270], label="rotate image (for smartphones)")
+            with gr.Column():
+              input_invert = gr.inputs.Checkbox(label="invert depthmap")
+        submit = gr.Button('generate image')
+
+    with gr.Column():
+        gr.Markdown('''
+          ## Outputs...
+        ''')
+        output_image = gr.Image(label="output image")
+        with gr.Accordion(label="depth map image", open=False):
+          depth_map = gr.Image(label="depth map")
+        output_text = gr.Textbox(label="output details")
+
+  submit.click(fn=controlnet_function, inputs=[input_prompt, input_image, input_negative_prompt, input_seed, input_rotate, input_invert], outputs=[output_text, output_image, depth_map])
+
+# depth_controlnet_gradio.launch(debug=False)
+
+
+
+
+
 # ----- Launch Tabs -----------------------------------------------------------------
 
-tabbed_interface = gr.TabbedInterface([new_welcome, advanced_tab, beta, canny_blocks_interface], ["Artbots", "Advanced", "Beta", "ControlNet"])
+tabbed_interface = gr.TabbedInterface([new_welcome, advanced_tab, beta, canny_blocks_interface, depth_controlnet_gradio], ["Welcome", "Advanced", "Beta", "EdgeTrace", "DepthMap"])
 # tabbed_interface = gr.TabbedInterface([new_welcome, advanced_tab, beta], ["Artbots", "Advanced", "Beta"])
 tabbed_interface.launch()