convert into fastapi

app.py

@@ -1,123 +1,57 @@
-# ——————————————————————————————————————————————————————————————
-# Monkey‐patch for diffusers<=0.19.3 which still does
-#   from huggingface_hub import cached_download
-#
-# New HF-Hub versions (>=0.14.0) removed cached_download, so we alias it.
-# This must appear before any imports of diffusers / transformers.
-# ——————————————————————————————————————————————————————————————
 import huggingface_hub
 huggingface_hub.cached_download = huggingface_hub.hf_hub_download
-
-
-import spaces
+from huggingface_hub import hf_hub_download
 
 import os
-import imageio
+import io
+import base64
+import tempfile
 import numpy as np
 import torch
-import rembg
+from fastapi import FastAPI, File, UploadFile, Form, HTTPException
+from fastapi.responses import JSONResponse
 from PIL import Image
 from torchvision.transforms import v2
 from pytorch_lightning import seed_everything
 from omegaconf import OmegaConf
-from einops import rearrange, repeat
-from tqdm import tqdm
+from einops import rearrange
 from diffusers import DiffusionPipeline, EulerAncestralDiscreteScheduler
 
+# Monkey-patch for diffusers<=0.19.3 which still does
+#    from huggingface_hub import cached_download
+# New HF-Hub versions (>=0.14.0) removed cached_download, so we alias it.
+
+
+# your util functions & model loaders
 from src.utils.train_util import instantiate_from_config
 from src.utils.camera_util import (
-    FOV_to_intrinsics, 
+    FOV_to_intrinsics,
     get_zero123plus_input_cameras,
     get_circular_camera_poses,
 )
 from src.utils.mesh_util import save_obj, save_glb
-from src.utils.infer_util import remove_background, resize_foreground, images_to_video
-
-import tempfile
-from functools import partial
-
-from huggingface_hub import hf_hub_download
-
-import gradio as gr
-
-
-def get_render_cameras(batch_size=1, M=120, radius=2.5, elevation=10.0, is_flexicubes=False):
-    """
-    Get the rendering camera parameters.
-    """
-    c2ws = get_circular_camera_poses(M=M, radius=radius, elevation=elevation)
-    if is_flexicubes:
-        cameras = torch.linalg.inv(c2ws)
-        cameras = cameras.unsqueeze(0).repeat(batch_size, 1, 1, 1)
-    else:
-        extrinsics = c2ws.flatten(-2)
-        intrinsics = FOV_to_intrinsics(50.0).unsqueeze(0).repeat(M, 1, 1).float().flatten(-2)
-        cameras = torch.cat([extrinsics, intrinsics], dim=-1)
-        cameras = cameras.unsqueeze(0).repeat(batch_size, 1, 1)
-    return cameras
-
-
-def images_to_video(images, output_path, fps=30):
-    # images: (N, C, H, W)
-    os.makedirs(os.path.dirname(output_path), exist_ok=True)
-    frames = []
-    for i in range(images.shape[0]):
-        frame = (images[i].permute(1, 2, 0).cpu().numpy() * 255).astype(np.uint8).clip(0, 255)
-        assert frame.shape[0] == images.shape[2] and frame.shape[1] == images.shape[3], \
-            f"Frame shape mismatch: {frame.shape} vs {images.shape}"
-        assert frame.min() >= 0 and frame.max() <= 255, \
-            f"Frame value out of range: {frame.min()} ~ {frame.max()}"
-        frames.append(frame)
-    imageio.mimwrite(output_path, np.stack(frames), fps=fps, codec='h264')
-
+from src.utils.infer_util import remove_background, resize_foreground
 
-###############################################################################
-# Configuration.
-###############################################################################
-
-import shutil
-
-def find_cuda():
-    # Check if CUDA_HOME or CUDA_PATH environment variables are set
-    cuda_home = os.environ.get('CUDA_HOME') or os.environ.get('CUDA_PATH')
-
-    if cuda_home and os.path.exists(cuda_home):
-        return cuda_home
-
-    # Search for the nvcc executable in the system's PATH
-    nvcc_path = shutil.which('nvcc')
-
-    if nvcc_path:
-        # Remove the 'bin/nvcc' part to get the CUDA installation path
-        cuda_path = os.path.dirname(os.path.dirname(nvcc_path))
-        return cuda_path
-
-    return None
-
-cuda_path = find_cuda()
-
-if cuda_path:
-    print(f"CUDA installation found at: {cuda_path}")
-else:
-    print("CUDA installation not found")
+# —————————————————————————————————————————————————————
+# 1) CONFIGURATION & MODEL LOADING
+# —————————————————————————————————————————————————————
 
+# Load our YAML config
 config_path = 'configs/instant-mesh-large.yaml'
 config = OmegaConf.load(config_path)
-config_name = os.path.basename(config_path).replace('.yaml', '')
 model_config = config.model_config
 infer_config = config.infer_config
+IS_FLEXICUBES = os.path.basename(config_path).startswith('instant-mesh')
 
-IS_FLEXICUBES = True if config_name.startswith('instant-mesh') else False
-
-# pick GPU if available, else CPU
+# pick device
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 if device.type == 'cpu':
-    print("⚠️  No CUDA GPU detected. Falling back to CPU (this will be very slow!)")
+    print("⚠️  No CUDA GPU detected. Falling back to CPU (very slow!)")
 
-# load diffusion model
-print('Loading diffusion model ...')
+# ——— Load diffusion (Zero123) pipeline ———
+print("Loading diffusion model …")
 pipeline = DiffusionPipeline.from_pretrained(
-    "sudo-ai/zero123plus-v1.2", 
+    "sudo-ai/zero123plus-v1.2",
     custom_pipeline="zero123plus",
     torch_dtype=torch.float16,
 )
@@ -125,280 +59,163 @@ pipeline.scheduler = EulerAncestralDiscreteScheduler.from_config(
     pipeline.scheduler.config, timestep_spacing='trailing'
 )
 
-# load custom white-background UNet
-unet_ckpt_path = hf_hub_download(repo_id="TencentARC/InstantMesh", filename="diffusion_pytorch_model.bin", repo_type="model")
-state_dict = torch.load(unet_ckpt_path, map_location='cpu')
-pipeline.unet.load_state_dict(state_dict, strict=True)
-
+# patch UNet to white-background version
+unet_ckpt = hf_hub_download(
+    repo_id="TencentARC/InstantMesh",
+    filename="diffusion_pytorch_model.bin",
+    repo_type="model",
+)
+sd = torch.load(unet_ckpt, map_location='cpu')
+pipeline.unet.load_state_dict(sd, strict=True)
 pipeline = pipeline.to(device)
 
-# load reconstruction model
-print('Loading reconstruction model ...')
-model_ckpt_path = hf_hub_download(repo_id="TencentARC/InstantMesh", filename="instant_mesh_large.ckpt", repo_type="model")
+# ——— Load reconstruction (InstantMesh) model ———
+print("Loading reconstruction model …")
+model_ckpt = hf_hub_download(
+    repo_id="TencentARC/InstantMesh",
+    filename="instant_mesh_large.ckpt",
+    repo_type="model",
+)
 model = instantiate_from_config(model_config)
-state_dict = torch.load(model_ckpt_path, map_location='cpu')['state_dict']
-state_dict = {k[14:]: v for k, v in state_dict.items() if k.startswith('lrm_generator.') and 'source_camera' not in k}
-model.load_state_dict(state_dict, strict=True)
-
-model = model.to(device)
-
-print('Loading Finished!')
-
-
-def check_input_image(input_image):
-    if input_image is None:
-        raise gr.Error("No image uploaded!")
-
+full_sd = torch.load(model_ckpt, map_location='cpu')['state_dict']
+# strip the "lrm_generator." prefix & unwanted keys
+sd = {
+    k[len("lrm_generator.") :]: v
+    for k, v in full_sd.items()
+    if k.startswith("lrm_generator.") and "source_camera" not in k
+}
+model.load_state_dict(sd, strict=True)
+model = model.to(device).eval()
+print("Models loaded ✅")
 
-def preprocess(input_image, do_remove_background):
+# —————————————————————————————————————————————————————
+# 2) HELPERS & INFERENCE LOGIC
+# —————————————————————————————————————————————————————
 
-    rembg_session = rembg.new_session() if do_remove_background else None
+def get_render_cameras(batch_size=1, M=120, radius=2.5, elevation=10.0, is_flexicubes=False):
+    c2ws = get_circular_camera_poses(M=M, radius=radius, elevation=elevation)
+    if is_flexicubes:
+        cams = torch.linalg.inv(c2ws)
+        return cams.unsqueeze(0).repeat(batch_size, 1, 1, 1)
+    else:
+        ext = c2ws.flatten(-2)
+        intr = FOV_to_intrinsics(50.0).unsqueeze(0).repeat(M, 1, 1).float().flatten(-2)
+        cams = torch.cat([ext, intr], dim=-1)
+        return cams.unsqueeze(0).repeat(batch_size, 1, 1)
 
+def preprocess(input_image: Image.Image, do_remove_background: bool):
+    rembg_sess = None
     if do_remove_background:
-        input_image = remove_background(input_image, rembg_session)
+        rembg_sess = __import__("rembg").new_session()
+        input_image = remove_background(input_image, rembg_sess)
         input_image = resize_foreground(input_image, 0.85)
-
     return input_image
 
-
-@spaces.GPU
-def generate_mvs(input_image, sample_steps, sample_seed):
-
+def generate_mvs(
+    input_image: Image.Image, sample_steps: int, sample_seed: int
+) -> tuple[Image.Image, Image.Image]:
+    """Return (raw_multi_view, preview_image)."""
     seed_everything(sample_seed)
-    
-    # sampling
-    z123_image = pipeline(
-        input_image, 
-        num_inference_steps=sample_steps
-    ).images[0]
-
-    show_image = np.asarray(z123_image, dtype=np.uint8)
-    show_image = torch.from_numpy(show_image)     # (960, 640, 3)
-    show_image = rearrange(show_image, '(n h) (m w) c -> (n m) h w c', n=3, m=2)
-    show_image = rearrange(show_image, '(n m) h w c -> (n h) (m w) c', n=2, m=3)
-    show_image = Image.fromarray(show_image.numpy())
-
-    return z123_image, show_image
-
-
-@spaces.GPU
-def make3d(images):
-
-    global model
+    out = pipeline(input_image, num_inference_steps=sample_steps)
+    mv = out.images[0]  # PIL, shape (H×W×3)
+
+    # create a tiled preview
+    arr = np.asarray(mv, dtype=np.uint8)
+    t = torch.from_numpy(arr)
+    t = rearrange(t, "(n h) (m w) c -> (n m) h w c", n=3, m=2)
+    t = rearrange(t, "(n m) h w c -> (n h) (m w) c", n=2, m=3)
+    preview = Image.fromarray(t.numpy())
+    return mv, preview
+
+def make3d(
+    mv: Image.Image,
+) -> tuple[str, str]:
+    """Return (path_to_obj, path_to_glb)."""
+    # initialize flexicubes if needed
     if IS_FLEXICUBES:
         model.init_flexicubes_geometry(device, use_renderer=False)
-    model = model.eval()
 
-    images = np.asarray(images, dtype=np.float32) / 255.0
-    images = torch.from_numpy(images).permute(2, 0, 1).contiguous().float()     # (3, 960, 640)
-    images = rearrange(images, 'c (n h) (m w) -> (n m) c h w', n=3, m=2)        # (6, 3, 320, 320)
+    # normalize & reshape
+    imgs = np.asarray(mv, dtype=np.float32) / 255.0
+    t = torch.from_numpy(imgs).permute(2, 0, 1).contiguous().float()
+    t = rearrange(t, "c (n h) (m w) -> (n m) c h w", n=3, m=2)
 
-    input_cameras = get_zero123plus_input_cameras(batch_size=1, radius=4.0).to(device)
-    render_cameras = get_render_cameras(batch_size=1, radius=2.5, is_flexicubes=IS_FLEXICUBES).to(device)
+    cam_in = get_zero123plus_input_cameras(1, radius=4.0).to(device)
+    cam_render = get_render_cameras(
+        1, radius=2.5, is_flexicubes=IS_FLEXICUBES
+    ).to(device)
 
-    images = images.unsqueeze(0).to(device)
-    images = v2.functional.resize(images, (320, 320), interpolation=3, antialias=True).clamp(0, 1)
+    t = t.unsqueeze(0).to(device)
+    t = v2.functional.resize(t, (320, 320), interpolation=3, antialias=True).clamp(0, 1)
 
-    mesh_fpath = tempfile.NamedTemporaryFile(suffix=f".obj", delete=False).name
-    print(mesh_fpath)
-    mesh_basename = os.path.basename(mesh_fpath).split('.')[0]
-    mesh_dirname = os.path.dirname(mesh_fpath)
-    video_fpath = os.path.join(mesh_dirname, f"{mesh_basename}.mp4")
-    mesh_glb_fpath = os.path.join(mesh_dirname, f"{mesh_basename}.glb")
+    # temp file names
+    obj_f = tempfile.NamedTemporaryFile(suffix=".obj", delete=False).name
+    base = os.path.splitext(obj_f)[0]
+    glb_f = base + ".glb"
 
     with torch.no_grad():
-        # get triplane
-        planes = model.forward_planes(images, input_cameras)
-
-        # # get video
-        # chunk_size = 20 if IS_FLEXICUBES else 1
-        # render_size = 384
-        
-        # frames = []
-        # for i in tqdm(range(0, render_cameras.shape[1], chunk_size)):
-        #     if IS_FLEXICUBES:
-        #         frame = model.forward_geometry(
-        #             planes,
-        #             render_cameras[:, i:i+chunk_size],
-        #             render_size=render_size,
-        #         )['img']
-        #     else:
-        #         frame = model.synthesizer(
-        #             planes,
-        #             cameras=render_cameras[:, i:i+chunk_size],
-        #             render_size=render_size,
-        #         )['images_rgb']
-        #     frames.append(frame)
-        # frames = torch.cat(frames, dim=1)
-
-        # images_to_video(
-        #     frames[0],
-        #     video_fpath,
-        #     fps=30,
-        # )
-
-        # print(f"Video saved to {video_fpath}")
-
-        # get mesh
-        mesh_out = model.extract_mesh(
-            planes,
-            use_texture_map=False,
-            **infer_config,
+        planes = model.forward_planes(t, cam_in)
+        mesh = model.extract_mesh(
+            planes, use_texture_map=False, **infer_config
         )
-
-        vertices, faces, vertex_colors = mesh_out
-        vertices = vertices[:, [1, 2, 0]]
-        
-        save_glb(vertices, faces, vertex_colors, mesh_glb_fpath)
-        save_obj(vertices, faces, vertex_colors, mesh_fpath)
-        
-        print(f"Mesh saved to {mesh_fpath}")
-
-    return mesh_fpath, mesh_glb_fpath
-
-
-_HEADER_ = '''
-<h2><b>Official 🤗 Gradio Demo</b></h2><h2><a href='https://github.com/TencentARC/InstantMesh' target='_blank'><b>InstantMesh: Efficient 3D Mesh Generation from a Single Image with Sparse-view Large Reconstruction Models</b></a></h2>
-
-**InstantMesh** is a feed-forward framework for efficient 3D mesh generation from a single image based on the LRM/Instant3D architecture.
-
-Code: <a href='https://github.com/TencentARC/InstantMesh' target='_blank'>GitHub</a>. Techenical report: <a href='https://arxiv.org/abs/2404.07191' target='_blank'>ArXiv</a>.
-
-❗️❗️❗️**Important Notes:**
-- Our demo can export a .obj mesh with vertex colors or a .glb mesh now. If you prefer to export a .obj mesh with a **texture map**, please refer to our <a href='https://github.com/TencentARC/InstantMesh?tab=readme-ov-file#running-with-command-line' target='_blank'>Github Repo</a>.
-- The 3D mesh generation results highly depend on the quality of generated multi-view images. Please try a different **seed value** if the result is unsatisfying (Default: 42).
-'''
-
-_CITE_ = r"""
-If InstantMesh is helpful, please help to ⭐ the <a href='https://github.com/TencentARC/InstantMesh' target='_blank'>Github Repo</a>. Thanks! [![GitHub Stars](https://img.shields.io/github/stars/TencentARC/InstantMesh?style=social)](https://github.com/TencentARC/InstantMesh)
----
-📝 **Citation**
-
-If you find our work useful for your research or applications, please cite using this bibtex:
-```bibtex
-@article{xu2024instantmesh,
-  title={InstantMesh: Efficient 3D Mesh Generation from a Single Image with Sparse-view Large Reconstruction Models},
-  author={Xu, Jiale and Cheng, Weihao and Gao, Yiming and Wang, Xintao and Gao, Shenghua and Shan, Ying},
-  journal={arXiv preprint arXiv:2404.07191},
-  year={2024}
-}
-```
-
-📋 **License**
-
-Apache-2.0 LICENSE. Please refer to the [LICENSE file](https://huggingface.co/spaces/TencentARC/InstantMesh/blob/main/LICENSE) for details.
-
-📧 **Contact**
-
-If you have any questions, feel free to open a discussion or contact us at <b>bluestyle928@gmail.com</b>.
-"""
-
-
-with gr.Blocks() as demo:
-    gr.Markdown(_HEADER_)
-    with gr.Row(variant="panel"):
-        with gr.Column():
-            with gr.Row():
-                input_image = gr.Image(
-                    label="Input Image",
-                    image_mode="RGBA",
-                    sources="upload",
-                    #width=256,
-                    #height=256,
-                    type="pil",
-                    elem_id="content_image",
-                )
-                processed_image = gr.Image(
-                    label="Processed Image", 
-                    image_mode="RGBA", 
-                    #width=256,
-                    #height=256,
-                    type="pil", 
-                    interactive=False
-                )
-            with gr.Row():
-                with gr.Group():
-                    do_remove_background = gr.Checkbox(
-                        label="Remove Background", value=True
-                    )
-                    sample_seed = gr.Number(value=42, label="Seed Value", precision=0)
-
-                    sample_steps = gr.Slider(
-                        label="Sample Steps",
-                        minimum=30,
-                        maximum=75,
-                        value=75,
-                        step=5
-                    )
-
-            with gr.Row():
-                submit = gr.Button("Generate", elem_id="generate", variant="primary")
-
-            with gr.Row(variant="panel"):
-                gr.Examples(
-                    examples=[
-                        os.path.join("examples", img_name) for img_name in sorted(os.listdir("examples"))
-                    ],
-                    inputs=[input_image],
-                    label="Examples",
-                    cache_examples=False,
-                    examples_per_page=16
-                )
-
-        with gr.Column():
-
-            with gr.Row():
-
-                with gr.Column():
-                    mv_show_images = gr.Image(
-                        label="Generated Multi-views",
-                        type="pil",
-                        width=379,
-                        interactive=False
-                    )
-
-                # with gr.Column():
-                #     output_video = gr.Video(
-                #         label="video", format="mp4",
-                #         width=379,
-                #         autoplay=True,
-                #         interactive=False
-                #     )
-
-            with gr.Row():
-                with gr.Tab("OBJ"):
-                    output_model_obj = gr.Model3D(
-                        label="Output Model (OBJ Format)",
-                        interactive=False,
-                    )
-                    gr.Markdown("Note: Downloaded .obj model will be flipped. Export .glb instead or manually flip it before usage.")
-                with gr.Tab("GLB"):
-                    output_model_glb = gr.Model3D(
-                        label="Output Model (GLB Format)",
-                        interactive=False,
-                    )
-                    gr.Markdown("Note: The model shown here has a darker appearance. Download to get correct results.")
-
-            with gr.Row():
-                gr.Markdown('''Try a different <b>seed value</b> if the result is unsatisfying (Default: 42).''')
-
-    gr.Markdown(_CITE_)
-
-    mv_images = gr.State()
-
-    submit.click(fn=check_input_image, inputs=[input_image]).success(
-        fn=preprocess,
-        inputs=[input_image, do_remove_background],
-        outputs=[processed_image],
-    ).success(
-        fn=generate_mvs,
-        inputs=[processed_image, sample_steps, sample_seed],
-        outputs=[mv_images, mv_show_images]
-        
-    ).success(
-        fn=make3d,
-        inputs=[mv_images],
-        outputs=[output_model_obj, output_model_glb]
+        verts, faces, colors = mesh
+        verts = verts[:, [1, 2, 0]]
+        save_obj(verts, faces, colors, obj_f)
+        save_glb(verts, faces, colors, glb_f)
+
+    return obj_f, glb_f
+
+def _pil_to_b64(img: Image.Image, fmt: str = "PNG") -> str:
+    buf = io.BytesIO()
+    img.save(buf, fmt)
+    return base64.b64encode(buf.getvalue()).decode()
+
+# —————————————————————————————————————————————————————
+# 3) FASTAPI APP
+# —————————————————————————————————————————————————————
+
+app = FastAPI(title="InstantMesh FastAPI Demo")
+
+@app.post("/infer")
+async def infer(
+    file: UploadFile = File(...),
+    remove_background: bool = Form(True),
+    sample_steps: int = Form(75, ge=1, le=100),
+    sample_seed: int = Form(42),
+):
+    # 1) load the RGBA image
+    data = await file.read()
+    try:
+        img = Image.open(io.BytesIO(data)).convert("RGBA")
+    except Exception:
+        raise HTTPException(400, detail="Invalid image")
+
+    # 2) run through pipeline
+    proc = preprocess(img, remove_background)
+    mv_raw, mv_preview = generate_mvs(proc, sample_steps, sample_seed)
+    obj_path, glb_path = make3d(mv_raw)
+
+    # 3) read back the mesh bytes
+    with open(obj_path, "rb") as f:
+        obj_b = f.read()
+    with open(glb_path, "rb") as f:
+        glb_b = f.read()
+
+    return JSONResponse(
+        {
+            "preview_png": _pil_to_b64(mv_preview),
+            "multi_views_png": _pil_to_b64(mv_raw),
+            "obj_data_b64": base64.b64encode(obj_b).decode(),
+            "glb_data_b64": base64.b64encode(glb_b).decode(),
+        }
     )
 
-demo.launch()
+if __name__ == "__main__":
+    import uvicorn
+
+    uvicorn.run(
+        "main:app",
+        host="0.0.0.0",
+        port=int(os.environ.get("PORT", 8000)),
+        reload=True,
+    )

requirements.txt

@@ -21,4 +21,5 @@ plyfile
 xformers==0.0.22.post7
 git+https://github.com/NVlabs/nvdiffrast/
 huggingface-hub
-onnxruntime
+onnxruntime
+fastapi