rodin

README.md

@@ -1,10 +1,10 @@
 ---
-title: Rodin
-emoji: 🐠
-colorFrom: yellow
-colorTo: indigo
+title: RodinTest
+emoji: 🌖
+colorFrom: green
+colorTo: green
 sdk: gradio
-sdk_version: 4.31.5
+sdk_version: 4.31.2
 app_file: app.py
 pinned: false
 ---

Rodin.py

@@ -0,0 +1,193 @@
+import socketio
+import requests
+import json
+import random
+import base64
+import io
+from PIL import Image
+from requests_toolbelt.multipart.encoder import MultipartEncoder
+from constant import *
+
+
+def login(email, password):
+    payload = {'password': password}
+    if email:
+        payload['email'] = email
+    
+    response = requests.post(f"{BASE_URL}/user/login", json=payload)
+    response_data = response.json()
+    
+    if 'error' in response_data and response_data['error']:
+        raise Exception(response_data['error'])
+    print("Login successful")
+    user_uuid = response_data['user_uuid']
+    token = response_data['token']
+    
+    return user_uuid, token
+
+def rodin_history(task_uuid, token):
+    headers = {
+        'Authorization': f'Bearer {token}'
+    }
+    response = requests.post(f"{BASE_URL}/task/rodin_history", data={"uuid": task_uuid}, headers=headers)
+    return response.json()
+
+def rodin_preprocess_image(generate_prompt, image, name, token):
+    m = MultipartEncoder(
+        fields={
+            'generate_prompt': "true" if generate_prompt else "false",
+            'images': (name, image, 'image/jpeg')
+        }
+    )
+    headers = {
+        'Content-Type': m.content_type,
+        'Authorization': f'Bearer {token}'
+    }
+    response = requests.post(f"{BASE_URL}/task/rodin_mesh_image_process", data=m, headers=headers)
+    return response.json()
+
+def crop_image(image, type):
+    new_image_width = 360 * (11520 // 720)  # 每隔720像素裁切一次，每次裁切宽度为360
+    new_image_height = 360  # 新图片的高度
+    new_image = Image.new('RGB', (new_image_width, new_image_height))
+
+    for i in range(11520 // 720):
+        left = i * 720 + type[1]
+        upper = type[0]
+        right = left + 360
+        lower = upper + 360
+        cropped_image = image.crop((left, upper, right, lower))
+        new_image.paste(cropped_image, (i * 360, 0))
+    return new_image
+
+# Perform Rodin mesh operation
+def rodin_mesh(prompt, group_uuid, settings, images, name, token):
+    images = [convert_base64_to_binary(img) for img in images]
+    
+    m = MultipartEncoder(
+        fields={
+            'prompt': prompt,
+            'group_uuid': group_uuid,
+            'settings': json.dumps(settings),  # Convert settings dictionary to JSON string
+            **{f'images': (name, image, 'image/jpeg') for i, image in enumerate(images)}
+        }
+    )
+
+    headers = {
+        'Content-Type': m.content_type,
+        'Authorization': f'Bearer {token}'
+    }
+    response = requests.post(f"{BASE_URL}/task/rodin_mesh", data=m, headers=headers)
+    return response.json()
+
+# Convert base64 to binary since the result from `rodin_preprocess_image` is encoded with base64
+def convert_base64_to_binary(base64_string):
+    if ',' in base64_string:
+        base64_string = base64_string.split(',')[1]
+        
+    image_data = base64.b64decode(base64_string)
+    image_buffer = io.BytesIO(image_data)
+    
+    return image_buffer
+
+def rodin_update(prompt, task_uuid, token, settings):
+    headers = {
+        'Authorization': f'Bearer {token}'
+    }
+    response = requests.post(f"{BASE_URL}/task/rodin_update", data={"uuid": task_uuid, "prompt": prompt, "settings": settings}, headers=headers)
+    return response.json()
+
+
+class Generator:
+    def __init__(self, user_id, password) -> None:
+        _, self.token = login(user_id, password)
+        self.task_uuid = None
+        
+    def preprocess(self, prompt, image_path):
+        image_file = open(image_path, 'rb')
+
+        if image_file == None:
+            print("Invalid image file.")
+
+        try:
+            if not prompt:
+                preprocess_response = rodin_preprocess_image(generate_prompt=True, image=image_file, name="images.jpeg", token=self.token)
+            else:
+                preprocess_response = rodin_preprocess_image(generate_prompt=False, image=image_file, name="images.jpeg", token=self.token)
+            if 'error' in preprocess_response:
+                print("Error in image preprocessing:", preprocess_response['error'])
+            else:
+                if not prompt:
+                    prompt = preprocess_response.get('prompt', 'Default prompt if none returned')                
+                processed_image = "data:image/png;base64," + preprocess_response.get('processed_image', None)
+        finally:
+            image_file.close()
+
+        return prompt, processed_image
+    
+    def generate_mesh(self, prompt, processed_image, task_uuid=""):
+        if task_uuid == "":
+            settings = {'view_weights': [1]}  # Define weights as per your requirements, for multiple images, use multiple values, e,g [0.5, 0.5]
+            images = [processed_image]  # List of images, all the images should be processed first
+            
+            mesh_response = rodin_mesh(prompt=prompt, group_uuid=None, settings=settings, images=images, name="images.jpeg", token=self.token)
+            progress_checker = JobStatusChecker(BASE_URL, mesh_response['job']['subscription_key'])
+            progress_checker.start()
+            
+            task_uuid = mesh_response['uuid'] # The task_uuid should be same during whole generation process
+        else:
+            new_prompt = prompt
+            settings = {
+                "view_weights": [1],
+                "seed": random.randint(0, 10000), # Customize your seed here
+                "escore": 5.5, # Temprature
+            }
+
+            update_response = rodin_update(new_prompt, task_uuid, self.token, settings)
+
+            # Check progress
+            subscription_key = update_response['job']['subscription_key']
+            checker = JobStatusChecker(BASE_URL, subscription_key)
+            checker.start()
+            
+        preview_image = rodin_history(task_uuid, self.token)["v1"]["preview_image"]
+        response = requests.get(preview_image, stream=True)
+        if response.status_code == 200:
+            # 创建一个PIL Image对象
+            image = Image.open(response.raw)
+            # 在这里对image对象进行处理，如显示、保存等
+        else:
+            print(f"Can't get the preview image. Status code：{response.status_code}")
+            raise RuntimeError
+        response.close()
+        return image, task_uuid, crop_image(image, DEFAULT)
+
+
+class JobStatusChecker:
+    def __init__(self, base_url, subscription_key):
+        self.base_url = base_url
+        self.subscription_key = subscription_key
+        self.sio = socketio.Client(logger=True, engineio_logger=True)
+
+        @self.sio.event
+        def connect():
+            print("Connected to the server.")
+
+        @self.sio.event
+        def disconnect():
+            print("Disconnected from server.")
+
+        @self.sio.on('message', namespace='*')
+        def message(*args, **kwargs):
+            if len(args) > 2:
+                data = args[2]
+                if data.get('jobStatus') == 'Succeeded':
+                    print("Job Succeeded! Please find the SDF image in history")
+                    self.sio.disconnect()
+            else:
+                print("Received event with insufficient arguments.")
+
+    def start(self):
+        self.sio.connect(f"{self.base_url}/scheduler_socket?subscription={self.subscription_key}", 
+                         namespaces=['/api/scheduler_socket'], transports='websocket')
+        self.sio.wait()

app.py

@@ -0,0 +1,205 @@
+import os
+# os.system('pip uninstall -y gradio_fake3d')
+# os.system('pip install gradio_fake3d-0.0.2-py3-none-any.whl')
+
+import gradio as gr
+import re
+from gradio_fake3d import Fake3D
+from PIL import Image
+from Rodin import Generator, crop_image
+from constant import *
+
+generator = Generator(USER, PASSWORD)
+
+change_button_name = """
+function updateButton(input) {
+    var buttonGenerate = document.getElementById('button_generate');
+    buttonGenerate.innerText = 'Redo';
+    return '';
+}
+"""
+
+reset_button_name = """
+function updateButton(input) {
+    var buttonGenerate = document.getElementById('button_generate');
+    buttonGenerate.innerText = 'Generate';
+    return '';
+}
+"""
+
+jump_to_rodin = """
+function redirectToGithub(input) {
+    if (input.includes('OpenClay')) {
+        window.open("https://github.com/CLAY-3D/OpenCLAY", "_blank");
+    }
+    return "Rodin Gen-1(0525)";
+}
+"""
+
+html_content = """
+<div style="text-align: center;">
+    <h1>Rodin Gen-1</h1>
+    <div style="display: flex; justify-content: space-around;">
+        <p><strong>Paper:</strong> <a href="https://sites.google.com/view/clay-3dlm" target="_blank"> https://sites.google.com/view/clay-3dlm</a></p>
+        <p><strong>Rodin Gen-1:</strong> <a href="https://hyperhuman.top/rodin" target="_blank">https://hyperhuman.top/rodin</a></p>
+        <p><strong>Github:</strong> <a href="https://github.com/CLAY-3D/OpenCLAY" target="_blank">https://github.com/CLAY-3D/OpenCLAY</a></p>
+    </div>
+</div>
+"""
+
+options = [
+    "Rodin Gen-1(0525)",
+    "OpenClay(600M) - Coming soon",
+    "OpenClay(200M) - Coming soon"
+]
+
+def do_nothing(text):
+    return ""
+
+def handle_selection(selection):
+    return "Rodin Gen-1(0525)"
+    # if selection in ["OpenClay(600M)", "OpenClay(200M)"]:
+    #     # 返回一个 HTML 字符串，用于在新窗口中打开指定的 URL
+    #     return f"<script>window.open('https://github.com/CLAY-3D/OpenCLAY', '_blank');</script>"
+    # else:
+    #     return "You selected Rodin Gen-1(0525)."
+
+def hint_in_prompt(hint, prompt):
+    return re.search(fr"{hint[:-1]}", prompt) is not None
+
+def prompt_remove_hint(prompt, hint):
+    return re.sub(fr"\s*{hint[:-1]}[\.,]*", "", prompt)
+
+def handle_hint_change(prompt: str, prompt_hint):
+    prompt = prompt.strip()
+    if prompt != "" and not prompt.endswith("."):
+        prompt = prompt + "."
+    for _, hint in PROMPT_HINT_LIST:
+        if hint in prompt_hint:
+            if not hint_in_prompt(hint, prompt):
+                prompt = prompt + " " + hint
+        else:
+            prompt = prompt_remove_hint(prompt, hint)
+    prompt = prompt.strip()
+    return prompt
+
+def handle_prompt_change(prompt):
+    hint_list = []
+    for _, hint in PROMPT_HINT_LIST:
+        if hint_in_prompt(hint, prompt):
+            hint_list.append(hint)
+
+    return hint_list
+
+def clear_task_uuid():
+    return ""
+
+def return_render(image):
+    image = Image.fromarray(image)
+    return image, crop_image(image, DEFAULT)
+
+def crop_image_default(image):
+    return crop_image(image, DEFAULT)
+
+def crop_image_metal(image):
+    return crop_image(image, METAL)
+
+def crop_image_contrast(image):
+    return crop_image(image, CONTRAST)
+
+def crop_image_normal(image):
+    return crop_image(image, NORMAL)
+
+with gr.Blocks() as demo:
+    gr.HTML(html_content)
+    with gr.Row():
+        with gr.Column():
+            block_image = gr.Image(height=256, image_mode="RGB", sources="upload", elem_classes="elem_imageupload", type="filepath")
+            block_model_card = gr.Dropdown(choices=options, label="Model Card", value="Rodin Gen-1(0525)", interactive=True)
+                # block_image_scale = gr.Slider(minimum=0, maximum=1, value=1, label="scale", interactive=True)
+            # block_model_card = gr.Radio(
+            #             choices=["Rodin Gen-1(0525)", "OpenClay(600M)", "OpenClay(200M)"],
+            #             label="Model Selection",
+            #             value="Rodin Gen-1(0525)"  # 默认选中第一个选项
+            #         )
+            with gr.Group():
+                # with gr.Row(equal_height=True):
+                block_prompt = gr.Textbox(
+                    value="",
+                    placeholder="Auto generated description of 3d geometry",
+                    lines=1,
+                    show_label=True,
+                    label="Prompt",
+                )
+                block_prompt_hint = gr.CheckboxGroup(value="Labels", choices=PROMPT_HINT_LIST)
+
+        with gr.Column():
+            with gr.Group():
+                fake3d = Fake3D(interactive=False, label="3D Preview")
+                with gr.Row():
+                    button_generate = gr.Button(value="Generate", variant="primary", elem_id="button_generate")
+
+                    with gr.Column(min_width=200, scale=20):
+                        with gr.Row():
+                            block_default = gr.Button("Default", min_width=0)
+                            block_metal = gr.Button("Metal", min_width=0)
+                        with gr.Row():
+                            block_contrast = gr.Button("Contrast", min_width=0)
+                            block_normal = gr.Button("Normal", min_width=0)
+            
+            button_more = gr.Button(value="Download", variant="primary", link=rodin_url)
+
+    cache_raw_image = gr.Image(visible=False, type="pil")
+    cache_image_base64 = gr.Text(visible=False)
+    cacha_empty = gr.Text(visible=False)
+    cache_task_uuid = gr.Text(value="", visible=False)
+
+    # button_generate.click(fn=return_render, inputs=[block_image], outputs=[raw_image, fake3d])
+    block_image.change(
+        fn=do_nothing, 
+        js=reset_button_name, 
+        inputs=[cacha_empty],
+        outputs=[cacha_empty]
+    ).then(fn=clear_task_uuid, outputs=[cache_task_uuid], show_progress="hidden")
+    
+    button_generate.click(
+        fn=generator.preprocess, 
+        inputs=[block_prompt, block_image], 
+        outputs=[block_prompt, cache_image_base64],
+        show_progress="minimal"
+    ).success(
+        fn=generator.generate_mesh,
+        inputs=[block_prompt, cache_image_base64, cache_task_uuid],
+        outputs=[cache_raw_image, cache_task_uuid, fake3d],
+    ).success(
+        fn=do_nothing, 
+        js=change_button_name, 
+        inputs=[cacha_empty],
+        outputs=[cacha_empty]
+    )
+    block_default.click(fn=crop_image_default, inputs=[cache_raw_image], outputs=fake3d, show_progress="minimal")
+    block_metal.click(fn=crop_image_metal, inputs=[cache_raw_image], outputs=fake3d, show_progress="minimal")
+    block_contrast.click(fn=crop_image_contrast, inputs=[cache_raw_image], outputs=fake3d, show_progress="minimal")
+    block_normal.click(fn=crop_image_normal, inputs=[cache_raw_image], outputs=fake3d, show_progress="minimal")
+    
+    button_more.click()
+    
+    block_prompt_hint.input(
+        fn=handle_hint_change, inputs=[block_prompt, block_prompt_hint], outputs=[block_prompt],
+        show_progress="hidden",
+        queue=False,
+    )
+    
+    block_prompt.change(
+        fn=handle_prompt_change,
+        inputs=[block_prompt],
+        outputs=[block_prompt_hint],
+        trigger_mode="always_last",
+        show_progress="hidden",
+    )
+    
+    block_model_card.change(fn=handle_selection, inputs=[block_model_card], outputs=[block_model_card], show_progress="hidden", js=jump_to_rodin)
+
+
+if __name__ == "__main__":
+    demo.launch(show_api=False)

change_button_text.js

@@ -0,0 +1,12 @@
+console.log('DOMContentLoaded');
+document.addEventListener('DOMContentLoaded', function() {
+    var buttonGenerate = document.getElementById('button_generate');
+    var clickedOnce = false;  // 用来追踪按钮是否被点击过一次
+
+    buttonGenerate.addEventListener('click', function() {
+        if (!clickedOnce) {
+            this.innerText = 'Redo';
+            clickedOnce = true;  // 更新状态为已点击
+        }
+    });
+});

constant.py

@@ -0,0 +1,25 @@
+# Replace with your own backend
+BASE_URL = "https://hyperhuman.deemos.com/api"
+
+# Creds
+USER = "merife6322@duiter.com"
+PASSWORD = "lrhswjj1314"
+
+DEFAULT = [0, 0]
+CONTRAST = [360, 0]
+METAL = [0, 360]
+NORMAL = [360, 360]
+rodin_url = "https://hyperhuman.deemos.com/rodin"
+
+PROMPT_HINT_LIST = [
+    ["symmetric", "symmetric geometry."],
+    ["character", "character."],
+    ["low-poly", "low-poly."],
+    ["high-poly", "high-poly."],
+    ["sharp", "sharp edges."],
+    ["smooth", "smooth edges."],
+    ["simple", "simple geometry."],
+    ["complex", "complex geometry."],
+    ["single", "single asset."],
+    ["game-ready", "game-ready."],
+]

requirements.txt

@@ -0,0 +1,6 @@
+python-socketio
+requests
+pillow
+gradio==4.31.2
+requests-toolbelt
+websocket-client