Update app.py

app.py

@@ -19,17 +19,16 @@ from safetensors.torch import load_file
 from torchvision.transforms import v2
 from tqdm import tqdm
 
-from src.utils.camera_util import (FOV_to_intrinsics, get_circular_camera_poses, get_zero123plus_input_cameras)
+from src.utils.camera_util import (FOV_to_intrinsics, get_circular_camera_poses,
+                                   get_zero123plus_input_cameras)
 from src.utils.infer_util import (remove_background, resize_foreground)
-from src.utils.mesh_util import save_glb, save_obj, save_texmap
+from src.utils.mesh_util import save_glb, save_obj, save_obj_with_mtl
 from src.utils.train_util import instantiate_from_config
 
-# Inicializa un tensor en CUDA y verifica el dispositivo.
 zero = torch.Tensor([0]).cuda()
-print(zero.device) # Verifica que el dispositivo sea CUDA.
-print(zero.device) # Verifica nuevamente que el dispositivo sea CUDA.
+print(zero.device) # <-- 'cpu' 🤔
+print(zero.device) # <-- 'cuda:0' 🤗
 
-# Función para encontrar el path de CUDA.
 def find_cuda():
     cuda_home = os.environ.get('CUDA_HOME') or os.environ.get('CUDA_PATH')
     if cuda_home and os.path.exists(cuda_home):
@@ -42,7 +41,6 @@ def find_cuda():
 
     return None
 
-# Función para obtener las cámaras de renderizado.
 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:
@@ -50,17 +48,18 @@ def get_render_cameras(batch_size=1, M=120, radius=2.5, elevation=10.0, is_flexi
         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)
+        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
 
-# Verifica si la imagen de entrada es nula.
+
 def check_input_image(input_image):
     if input_image is None:
         raise gr.Error("No image selected!")
 
-# Preprocesa la imagen de entrada y opcionalmente elimina el fondo.
+
 def preprocess(input_image, do_remove_background):
     rembg_session = rembg.new_session() if do_remove_background else None
 
@@ -70,28 +69,31 @@ def preprocess(input_image, do_remove_background):
 
     return input_image
 
-# Genera vistas múltiples de la imagen de entrada.
+
 @spaces.GPU(duration=20)
 def generate_mvs(input_image, sample_steps, sample_seed):
     seed_everything(sample_seed)
     
-    print(zero.device) # Verifica que el dispositivo sea CUDA.
+    print(zero.device) # <-- 'cuda:0' 🤗
     
-    z123_image = pipeline(input_image, num_inference_steps=sample_steps).images[0]
+    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)
-    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 = 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
 
-# Convierte imágenes en modelos 3D y genera texturas.
+
 @spaces.GPU
 def make3d(images):
     
-    print(zero.device) # Verifica que el dispositivo sea CUDA.
+    print(zero.device) # <-- 'cuda:0' 🤗
     
     global model
     if IS_FLEXICUBES:
@@ -102,13 +104,16 @@ def make3d(images):
     images = torch.from_numpy(images).permute(2, 0, 1).contiguous().float()
     images = rearrange(images, '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)
+    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)
 
     images = images.unsqueeze(0).to(device)
-    images = v2.functional.resize(images, (320, 320), interpolation=3, antialias=True).clamp(0, 1)
+    images = v2.functional.resize(
+        images, (320, 320), interpolation=3, antialias=True).clamp(0, 1)
 
-    mesh_fpath = tempfile.NamedTemporaryFile(suffix=".obj", delete=False).name
+    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)
@@ -117,24 +122,38 @@ def make3d(images):
 
     with torch.no_grad():
         planes = model.forward_planes(images, input_cameras)
-        mesh_out = model.extract_mesh(planes, use_texture_map=True, **infer_config)
 
-        vertices, faces, vertex_colors, texture_map = mesh_out
-        vertices = vertices[:, [1, 2, 0]]
+        # Extraer el mesh y la textura
+        mesh_out = model.extract_mesh(
+            planes,
+            use_texture_map=True,  # Habilitar la generación de texturas
+            **infer_config,
+        )
+
+        # Guardar el mesh con o sin textura
+        if len(mesh_out) == 5:  # Si se genera una textura
+            vertices, faces, uvs, mesh_tex_idx, tex_map = mesh_out
+            save_obj_with_mtl(
+                vertices.data.cpu().numpy(),
+                uvs.data.cpu().numpy(),
+                faces.data.cpu().numpy(),
+                mesh_tex_idx.data.cpu().numpy(),
+                tex_map.permute(1, 2, 0).data.cpu().numpy(),
+                mesh_fpath,
+            )
+        else:
+            vertices, faces, vertex_colors = mesh_out
+            save_obj(vertices, faces, vertex_colors, mesh_fpath)
+
+        print(f"Mesh saved to {mesh_fpath}")
+
+    return mesh_fpath, mesh_glb_fpath, texmap_fpath if len(mesh_out) == 5 else None
 
-        save_glb(vertices, faces, vertex_colors, texture_map, mesh_glb_fpath)
-        save_obj(vertices, faces, vertex_colors, texture_map, mesh_fpath)
-        save_texmap(texture_map, texmap_fpath)
 
-        print(f"Mesh and texture saved to {mesh_fpath} and {texmap_fpath}")
-
-    return mesh_fpath, mesh_glb_fpath, texmap_fpath
-
-# Procesa la imagen generada a partir de un prompt de texto.
 @spaces.GPU
 def process_image(num_images, prompt):
     
-    print(zero.device) # Verifica que el dispositivo sea CUDA.
+    print(zero.device) # <-- 'cuda:0' 🤗
     
     global pipe
     with torch.inference_mode(), torch.autocast("cuda", dtype=torch.bfloat16):
@@ -148,7 +167,8 @@ def process_image(num_images, prompt):
             timesteps=[800]
         ).images
 
-# Configuración inicial del entorno CUDA y carga de configuración del modelo.
+
+# Configuration
 cuda_path = find_cuda()
 config_path = 'configs/instant-mesh-large.yaml'
 config = OmegaConf.load(config_path)
@@ -159,7 +179,7 @@ infer_config = config.infer_config
 IS_FLEXICUBES = config_name.startswith('instant-mesh')
 device = torch.device('cuda')
 
-# Carga del modelo de difusión.
+# Load diffusion model
 print('Loading diffusion model ...')
 pipeline = DiffusionPipeline.from_pretrained(
     "sudo-ai/zero123plus-v1.2",
@@ -177,20 +197,21 @@ pipeline.unet.load_state_dict(state_dict, strict=True)
 
 pipeline = pipeline.to(device)
 
-# Carga del modelo de reconstrucción.
+# 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")
 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}
+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('Carga Completa!')
 
-# Interfaz de usuario usando Gradio.
+# Gradio UI
 with gr.Blocks() as demo:
     with gr.Row(variant="panel"):
         with gr.Column():