Update app.py

app.py

@@ -1,33 +1,20 @@
+import gradio as gr
+from gradio_litmodel3d import LitModel3D
+
 import os
-import sys
-import subprocess
-import gc
 import shutil
-from typing import *
-
-# [AUTO-INSTALL] accelerate 라이브러리
-try:
-    import accelerate
-except ImportError:
-    subprocess.check_call([sys.executable, "-m", "pip", "install", "accelerate"])
-
-# [중요] OOM 방지를 위한 메모리 파편화 설정
-os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
 os.environ['SPCONV_ALGO'] = 'native'
-
+from typing import *
 import torch
-import torch.nn as nn
 import numpy as np
 import imageio
 from easydict import EasyDict as edict
 from PIL import Image
-import gradio as gr
-from gradio_litmodel3d import LitModel3D
-
 from trellis.pipelines import TrellisVGGTTo3DPipeline
 from trellis.representations import Gaussian, MeshExtractResult
 from trellis.utils import render_utils, postprocessing_utils
-from accelerate import dispatch_model, infer_auto_device_map
+
+
 
 MAX_SEED = np.iinfo(np.int32).max
 TMP_DIR = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'tmp')
@@ -37,18 +24,42 @@ def start_session(req: gr.Request):
     user_dir = os.path.join(TMP_DIR, str(req.session_hash))
     os.makedirs(user_dir, exist_ok=True)
     
+    
 def end_session(req: gr.Request):
     user_dir = os.path.join(TMP_DIR, str(req.session_hash))
-    if os.path.exists(user_dir):
-        shutil.rmtree(user_dir)
-    gc.collect()
-    torch.cuda.empty_cache()
+    shutil.rmtree(user_dir)
 
 def preprocess_image(image: Image.Image) -> Image.Image:
+    """
+    Preprocess the input image for 3D generation.
+    
+    This function is called when a user uploads an image or selects an example.
+    It applies background removal and other preprocessing steps necessary for
+    optimal 3D model generation.
+
+    Args:
+        image (Image.Image): The input image from the user
+
+    Returns:
+        Image.Image: The preprocessed image ready for 3D generation
+    """
     processed_image = pipeline.preprocess_image(image)
     return processed_image
 
 def preprocess_videos(video: str) -> List[Tuple[Image.Image, str]]:
+    """
+    Preprocess the input video for multi-image 3D generation.
+    
+    This function is called when a user uploads a video.
+    It extracts frames from the video and processes each frame to prepare them
+    for the multi-image 3D generation pipeline.
+    
+    Args:
+        video (str): The path to the input video file
+        
+    Returns:
+        List[Tuple[Image.Image, str]]: The list of preprocessed images ready for 3D generation
+    """
     vid = imageio.get_reader(video, 'ffmpeg')
     fps = vid.get_meta_data()['fps']
     images = []
@@ -63,10 +74,23 @@ def preprocess_videos(video: str) -> List[Tuple[Image.Image, str]]:
     return processed_images
 
 def preprocess_images(images: List[Tuple[Image.Image, str]]) -> List[Image.Image]:
+    """
+    Preprocess a list of input images for multi-image 3D generation.
+    
+    This function is called when users upload multiple images in the gallery.
+    It processes each image to prepare them for the multi-image 3D generation pipeline.
+    
+    Args:
+        images (List[Tuple[Image.Image, str]]): The input images from the gallery
+        
+    Returns:
+        List[Image.Image]: The preprocessed images ready for 3D generation
+    """
     images = [image[0] for image in images]
     processed_images = [pipeline.preprocess_image(image) for image in images]
     return processed_images
 
+
 def pack_state(gs: Gaussian, mesh: MeshExtractResult) -> dict:
     return {
         'gaussian': {
@@ -82,9 +106,9 @@ def pack_state(gs: Gaussian, mesh: MeshExtractResult) -> dict:
             'faces': mesh.faces.cpu().numpy(),
         },
     }
-
+    
+    
 def unpack_state(state: dict) -> Tuple[Gaussian, edict, str]:
-    device = 'cuda:0' if torch.cuda.is_available() else 'cpu'
     gs = Gaussian(
         aabb=state['gaussian']['aabb'],
         sh_degree=state['gaussian']['sh_degree'],
@@ -93,21 +117,37 @@ def unpack_state(state: dict) -> Tuple[Gaussian, edict, str]:
         opacity_bias=state['gaussian']['opacity_bias'],
         scaling_activation=state['gaussian']['scaling_activation'],
     )
-    gs._xyz = torch.tensor(state['gaussian']['_xyz'], device=device)
-    gs._features_dc = torch.tensor(state['gaussian']['_features_dc'], device=device)
-    gs._scaling = torch.tensor(state['gaussian']['_scaling'], device=device)
-    gs._rotation = torch.tensor(state['gaussian']['_rotation'], device=device)
-    gs._opacity = torch.tensor(state['gaussian']['_opacity'], device=device)
+    gs._xyz = torch.tensor(state['gaussian']['_xyz'], device='cuda')
+    gs._features_dc = torch.tensor(state['gaussian']['_features_dc'], device='cuda')
+    gs._scaling = torch.tensor(state['gaussian']['_scaling'], device='cuda')
+    gs._rotation = torch.tensor(state['gaussian']['_rotation'], device='cuda')
+    gs._opacity = torch.tensor(state['gaussian']['_opacity'], device='cuda')
     
     mesh = edict(
-        vertices=torch.tensor(state['mesh']['vertices'], device=device),
-        faces=torch.tensor(state['mesh']['faces'], device=device),
+        vertices=torch.tensor(state['mesh']['vertices'], device='cuda'),
+        faces=torch.tensor(state['mesh']['faces'], device='cuda'),
     )
+    
     return gs, mesh
 
+
 def get_seed(randomize_seed: bool, seed: int) -> int:
+    """
+    Get the random seed for generation.
+    
+    This function is called by the generate button to determine whether to use
+    a random seed or the user-specified seed value.
+    
+    Args:
+        randomize_seed (bool): Whether to generate a random seed
+        seed (int): The user-specified seed value
+        
+    Returns:
+        int: The seed to use for generation
+    """
     return np.random.randint(0, MAX_SEED) if randomize_seed else seed
 
+
 def generate_and_extract_glb(
     multiimages: List[Tuple[Image.Image, str]],
     seed: int,
@@ -120,35 +160,54 @@ def generate_and_extract_glb(
     texture_size: int,
     req: gr.Request,
 ) -> Tuple[dict, str, str, str]:
-    
-    gc.collect()
-    torch.cuda.empty_cache()
-    
+    """
+    Convert an image to a 3D model and extract GLB file.
+
+    Args:
+        image (Image.Image): The input image.
+        multiimages (List[Tuple[Image.Image, str]]): The input images in multi-image mode.
+        is_multiimage (bool): Whether is in multi-image mode.
+        seed (int): The random seed.
+        ss_guidance_strength (float): The guidance strength for sparse structure generation.
+        ss_sampling_steps (int): The number of sampling steps for sparse structure generation.
+        slat_guidance_strength (float): The guidance strength for structured latent generation.
+        slat_sampling_steps (int): The number of sampling steps for structured latent generation.
+        multiimage_algo (Literal["multidiffusion", "stochastic"]): The algorithm for multi-image generation.
+        mesh_simplify (float): The mesh simplification factor.
+        texture_size (int): The texture resolution.
+
+    Returns:
+        dict: The information of the generated 3D model.
+        str: The path to the video of the 3D model.
+        str: The path to the extracted GLB file.
+        str: The path to the extracted GLB file (for download).
+    """
     user_dir = os.path.join(TMP_DIR, str(req.session_hash))
     image_files = [image[0] for image in multiimages]
 
-    try:
-        # [중요] 추론 시 그래디언트 계산 끔 (메모리 절약)
-        with torch.no_grad():
-            outputs, _, _ = pipeline.run(
-                image=image_files,
-                seed=seed,
-                formats=["gaussian", "mesh"],
-                preprocess_image=False,
-                sparse_structure_sampler_params={
-                    "steps": ss_sampling_steps,
-                    "cfg_strength": ss_guidance_strength,
-                },
-                slat_sampler_params={
-                    "steps": slat_sampling_steps,
-                    "cfg_strength": slat_guidance_strength,
-                },
-                mode=multiimage_algo,
-            )
-    except Exception as e:
-        torch.cuda.empty_cache()
-        # 구체적인 에러 메시지 반환
-        raise RuntimeError(f"Generation Failed: {str(e)}")
+    # Generate 3D model
+    outputs, _, _ = pipeline.run(
+        image=image_files,
+        seed=seed,
+        formats=["gaussian", "mesh"],
+        preprocess_image=False,
+        sparse_structure_sampler_params={
+            "steps": ss_sampling_steps,
+            "cfg_strength": ss_guidance_strength,
+        },
+        slat_sampler_params={
+            "steps": slat_sampling_steps,
+            "cfg_strength": slat_guidance_strength,
+        },
+        mode=multiimage_algo,
+    )
+
+    # Render video
+    # import uuid
+    # output_id = str(uuid.uuid4())
+    # os.makedirs(f"{TMP_DIR}/{output_id}", exist_ok=True)
+    # video_path = f"{TMP_DIR}/{output_id}/preview.mp4"
+    # glb_path = f"{TMP_DIR}/{output_id}/mesh.glb"
 
     video = render_utils.render_video(outputs['gaussian'][0], num_frames=120)['color']
     video_geo = render_utils.render_video(outputs['mesh'][0], num_frames=120)['normal']
@@ -156,32 +215,44 @@ def generate_and_extract_glb(
     video_path = os.path.join(user_dir, 'sample.mp4')
     imageio.mimsave(video_path, video, fps=15)
     
+    # Extract GLB
     gs = outputs['gaussian'][0]
     mesh = outputs['mesh'][0]
     glb = postprocessing_utils.to_glb(gs, mesh, simplify=mesh_simplify, texture_size=texture_size, verbose=False)
     glb_path = os.path.join(user_dir, 'sample.glb')
     glb.export(glb_path)
     
+    # Pack state for optional Gaussian extraction
     state = pack_state(gs, mesh)
     
-    del outputs, gs, mesh, glb
-    gc.collect()
     torch.cuda.empty_cache()
-    
     return state, video_path, glb_path, glb_path
 
+
 def extract_gaussian(state: dict, req: gr.Request) -> Tuple[str, str]:
+    """
+    Extract a Gaussian splatting file from the generated 3D model.
+    
+    This function is called when the user clicks "Extract Gaussian" button.
+    It converts the 3D model state into a .ply file format containing
+    Gaussian splatting data for advanced 3D applications.
+
+    Args:
+        state (dict): The state of the generated 3D model containing Gaussian data
+        req (gr.Request): Gradio request object for session management
+
+    Returns:
+        Tuple[str, str]: Paths to the extracted Gaussian file (for display and download)
+    """
     user_dir = os.path.join(TMP_DIR, str(req.session_hash))
     gs, _ = unpack_state(state)
     gaussian_path = os.path.join(user_dir, 'sample.ply')
     gs.save_ply(gaussian_path)
-    del gs
     torch.cuda.empty_cache()
     return gaussian_path, gaussian_path
 
+
 def prepare_multi_example() -> List[Image.Image]:
-    if not os.path.exists("assets/example_multi_image"):
-        return []
     multi_case = list(set([i.split('_')[0] for i in os.listdir("assets/example_multi_image")]))
     images = []
     for case in multi_case:
@@ -196,7 +267,21 @@ def prepare_multi_example() -> List[Image.Image]:
             images.append(Image.fromarray(np.concatenate(_images, axis=1)))
     return images
 
+
 def split_image(image: Image.Image) -> List[Image.Image]:
+    """
+    Split a multi-view image into separate view images.
+    
+    This function is called when users select multi-image examples that contain
+    multiple views in a single concatenated image. It automatically splits them
+    based on alpha channel boundaries and preprocesses each view.
+    
+    Args:
+        image (Image.Image): A concatenated image containing multiple views
+        
+    Returns:
+        List[Image.Image]: List of individual preprocessed view images
+    """
     image = np.array(image)
     alpha = image[..., 3]
     alpha = np.any(alpha>0, axis=0)
@@ -219,7 +304,22 @@ demo = gr.Blocks(
     """
 )
 with demo:
-    gr.Markdown("# 💻 ReconViaGen (GPU 0 Freed)")
+    gr.Markdown("""
+    # 💻 ReconViaGen
+    <p align="center">
+    <a title="Github" href="https://github.com/GAP-LAB-CUHK-SZ/ReconViaGen" target="_blank" rel="noopener noreferrer" style="display: inline-block;">
+        <img src="https://img.shields.io/github/stars/GAP-LAB-CUHK-SZ/ReconViaGen?label=GitHub%20%E2%98%85&logo=github&color=C8C" alt="badge-github-stars">
+    </a>
+    <a title="Website" href="https://jiahao620.github.io/reconviagen/" target="_blank" rel="noopener noreferrer" style="display: inline-block;">
+        <img src="https://www.obukhov.ai/img/badges/badge-website.svg">
+    </a>
+    <a title="arXiv" href="https://jiahao620.github.io/reconviagen/" target="_blank" rel="noopener noreferrer" style="display: inline-block;">
+        <img src="https://www.obukhov.ai/img/badges/badge-pdf.svg">
+    </a>
+    </p>
+    
+    ✨This demo is partial. We will release the whole model later. Stay tuned!✨
+    """)
     
     with gr.Row():
         with gr.Column():
@@ -228,6 +328,9 @@ with demo:
                     input_video = gr.Video(label="Upload Video", interactive=True, height=300)
                     image_prompt = gr.Image(label="Image Prompt", format="png", visible=False, image_mode="RGBA", type="pil", height=300)
                     multiimage_prompt = gr.Gallery(label="Image Prompt", format="png", type="pil", height=300, columns=3)
+                    gr.Markdown("""
+                        Input different views of the object in separate images.
+                                """)
         
             with gr.Accordion(label="Generation Settings", open=False):
                 seed = gr.Slider(0, MAX_SEED, label="Seed", value=0, step=1)
@@ -248,6 +351,9 @@ with demo:
 
             generate_btn = gr.Button("Generate & Extract GLB", variant="primary")
             extract_gs_btn = gr.Button("Extract Gaussian", interactive=False)
+            gr.Markdown("""
+                        *NOTE: Gaussian file can be very large (~50MB), it will take a while to display and download.*
+                        """)
 
         with gr.Column():
             video_output = gr.Video(label="Generated 3D Asset", autoplay=True, loop=True, height=300)
@@ -259,6 +365,7 @@ with demo:
     
     output_buf = gr.State()
 
+    # Example images at the bottom of the page
     with gr.Row() as multiimage_example:
         examples_multi = gr.Examples(
             examples=prepare_multi_example(),
@@ -273,12 +380,25 @@ with demo:
     demo.load(start_session)
     demo.unload(end_session)
 
-    input_video.upload(preprocess_videos, inputs=[input_video], outputs=[multiimage_prompt])
-    input_video.clear(lambda: tuple([None, None]), outputs=[input_video, multiimage_prompt])
-    multiimage_prompt.upload(preprocess_images, inputs=[multiimage_prompt], outputs=[multiimage_prompt])
+    input_video.upload(
+        preprocess_videos,
+        inputs=[input_video],
+        outputs=[multiimage_prompt],
+    )
+    input_video.clear(
+        lambda: tuple([None, None]),
+        outputs=[input_video, multiimage_prompt],
+    )
+    multiimage_prompt.upload(
+        preprocess_images,
+        inputs=[multiimage_prompt],
+        outputs=[multiimage_prompt],
+    )
 
     generate_btn.click(
-        get_seed, inputs=[randomize_seed, seed], outputs=[seed]
+        get_seed,
+        inputs=[randomize_seed, seed],
+        outputs=[seed],
     ).then(
         generate_and_extract_glb,
         inputs=[multiimage_prompt, seed, ss_guidance_strength, ss_sampling_steps, slat_guidance_strength, slat_sampling_steps, multiimage_algo, mesh_simplify, texture_size],
@@ -293,59 +413,25 @@ with demo:
         outputs=[extract_gs_btn, download_glb, download_gs],
     )
     
-    extract_gs_btn.click(extract_gaussian, inputs=[output_buf], outputs=[model_output, download_gs]).then(
-        lambda: gr.Button(interactive=True), outputs=[download_gs]
+    extract_gs_btn.click(
+        extract_gaussian,
+        inputs=[output_buf],
+        outputs=[model_output, download_gs],
+    ).then(
+        lambda: gr.Button(interactive=True),
+        outputs=[download_gs],
     )
 
     model_output.clear(
         lambda: tuple([gr.Button(interactive=False), gr.Button(interactive=False)]),
         outputs=[download_glb, download_gs],
     )
+    
 
-# Launch Script
+# Launch the Gradio app
 if __name__ == "__main__":
-    print("🚀 Initializing Pipeline...")
-    # 1. Pipeline 로드
     pipeline = TrellisVGGTTo3DPipeline.from_pretrained("esther11/trellis-vggt-v0-2")
-    
-    # 2. 모든 모델을 일단 CUDA:0에 올려서 기본 설정(device mismatch 방지)을 완료함
     pipeline.cuda()
-    pipeline._device = torch.device("cuda:0") # 내부 device 속성 고정
-
-    gpu_count = torch.cuda.device_count()
-    print(f"⚡ Detected {gpu_count} GPUs.")
-
-    if gpu_count > 1:
-        print("⚡ Multi-GPU Mode: Offloading VGGT from GPU 0.")
-        
-        # [핵심 로직] GPU 0을 비우기 위한 전략
-        # VGGT 모델을 잠시 CPU로 내립니다.
-        pipeline.VGGT_model.cpu() 
-        
-        print("   - Calculating Device Map (Banning GPU 0 for VGGT)...")
-        
-        # max_memory 설정:
-        # GPU 0: "10MiB" (사실상 VGGT 모델 적재 금지)
-        # GPU 1~N: "20GiB" (여유롭게 할당)
-        max_mem = {0: "10MiB"} 
-        for i in range(1, gpu_count):
-            max_mem[i] = "20GiB"
-
-        # 이 설정으로 맵을 짜면 accelerate는 GPU 0을 건너뛰고 GPU 1부터 모델을 채웁니다.
-        device_map = infer_auto_device_map(
-            pipeline.VGGT_model, 
-            max_memory=max_mem, 
-            no_split_module_classes=["Block", "ResnetBlock"] 
-        )
-        
-        # 맵 적용하여 분산 로드
-        pipeline.VGGT_model = dispatch_model(pipeline.VGGT_model, device_map=device_map)
-        
-        print("✅ VGGT Model successfully pushed to GPU 1+.")
-        print("   - GPU 0: Birefnet (Preprocessing) + Controller")
-        print("   - GPU 1+: VGGT (Inference)")
-        
-    else:
-        print("⚠️ Warning: Only 1 GPU detected. Expect OOM if VRAM < 24GB.")
-
+    pipeline.VGGT_model.cuda()
+    pipeline.birefnet_model.cuda()
     demo.launch()