Update app.py

app.py

@@ -1,7 +1,10 @@
 import gradio as gr
+import spaces
+# 유료 환경에서는 spaces import가 있어도 @spaces.GPU 데코레이터만 안 쓰면 됩니다.
+from gradio_litmodel3d import LitModel3D
+
 import os
 import shutil
-
 os.environ['SPCONV_ALGO'] = 'native'
 from typing import *
 import torch
@@ -12,26 +15,54 @@ from PIL import Image
 from trellis.pipelines import TrellisVGGTTo3DPipeline
 from trellis.representations import Gaussian, MeshExtractResult
 from trellis.utils import render_utils, postprocessing_utils
-from gradio_litmodel3d import LitModel3D
+
+from wheels.vggt.vggt.utils.load_fn import load_and_preprocess_images
+from wheels.vggt.vggt.utils.pose_enc import pose_encoding_to_extri_intri
+import open3d as o3d
+from torchvision import transforms as TF
+from PIL import Image
+import sys
+# sys.path.append("wheels") # 필요시 경로 수정
+import cv2 # cv2가 누락되어 있을 수 있어 추가했습니다.
+
+from wheels.mast3r.model import AsymmetricMASt3R
+from wheels.mast3r.fast_nn import fast_reciprocal_NNs
+from wheels.dust3r.dust3r.inference import inference
+from wheels.dust3r.dust3r.utils.image import load_images_new
+from trellis.utils.general_utils import *
+import copy
 
 MAX_SEED = np.iinfo(np.int32).max
 TMP_DIR = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'tmp')
+# TMP_DIR = "tmp/Trellis-demo"
+# os.environ['GRADIO_TEMP_DIR'] = 'tmp'
 os.makedirs(TMP_DIR, exist_ok=True)
+device = "cuda" if torch.cuda.is_available() else "cpu"
 
 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))
+    # [수정] 폴더가 없으면 삭제하지 않도록 예외 처리 추가
     if os.path.exists(user_dir):
         shutil.rmtree(user_dir)
 
+# [수정] 유료 4 GPU 사용 시 충돌 방지를 위해 @spaces.GPU 제거
 def preprocess_image(image: Image.Image) -> Image.Image:
+    """
+    Preprocess the input image for 3D generation.
+    """
     processed_image = pipeline.preprocess_image(image)
     return processed_image
 
+# [수정] @spaces.GPU 제거
 def preprocess_videos(video: str) -> List[Tuple[Image.Image, str]]:
+    """
+    Preprocess the input video for multi-image 3D generation.
+    """
     vid = imageio.get_reader(video, 'ffmpeg')
     fps = vid.get_meta_data()['fps']
     images = []
@@ -45,11 +76,16 @@ def preprocess_videos(video: str) -> List[Tuple[Image.Image, str]]:
     processed_images = [pipeline.preprocess_image(image) for image in images]
     return processed_images
 
+# [수정] @spaces.GPU 제거
 def preprocess_images(images: List[Tuple[Image.Image, str]]) -> List[Image.Image]:
+    """
+    Preprocess a list of input images for multi-image 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': {
@@ -65,7 +101,8 @@ def pack_state(gs: Gaussian, mesh: MeshExtractResult) -> dict:
             'faces': mesh.faces.cpu().numpy(),
         },
     }
-
+    
+    
 def unpack_state(state: dict) -> Tuple[Gaussian, edict, str]:
     gs = Gaussian(
         aabb=state['gaussian']['aabb'],
@@ -75,7 +112,7 @@ def unpack_state(state: dict) -> Tuple[Gaussian, edict, str]:
         opacity_bias=state['gaussian']['opacity_bias'],
         scaling_activation=state['gaussian']['scaling_activation'],
     )
-    # 추론 시 메인 장치인 cuda:0으로 데이터를 보냅니다.
+    # [수정] 데이터를 로드할 때 메인 GPU(cuda:0)으로 보냄
     gs._xyz = torch.tensor(state['gaussian']['_xyz'], device='cuda:0')
     gs._features_dc = torch.tensor(state['gaussian']['_features_dc'], device='cuda:0')
     gs._scaling = torch.tensor(state['gaussian']['_scaling'], device='cuda:0')
@@ -86,11 +123,218 @@ def unpack_state(state: dict) -> Tuple[Gaussian, edict, str]:
         vertices=torch.tensor(state['mesh']['vertices'], device='cuda:0'),
         faces=torch.tensor(state['mesh']['faces'], device='cuda:0'),
     )
+    
     return gs, mesh
 
+
 def get_seed(randomize_seed: bool, seed: int) -> int:
+    """
+    Get the random seed for generation.
+    """
     return np.random.randint(0, MAX_SEED) if randomize_seed else seed
 
+def align_camera(num_frames, extrinsic, intrinsic, rend_extrinsics, rend_intrinsics):
+
+    extrinsic_tmp = extrinsic.clone()
+    camera_relative = torch.matmul(extrinsic_tmp[:num_frames,:3,:3].permute(0,2,1), extrinsic_tmp[num_frames:,:3,:3])
+    camera_relative_angle = torch.acos(((camera_relative[:,0,0] + camera_relative[:,1,1] + camera_relative[:,2,2] - 1) / 2).clamp(-1, 1))
+    idx = torch.argmin(camera_relative_angle)
+    target_extrinsic = rend_extrinsics[idx:idx+1].clone()
+
+    focal_x = intrinsic[:num_frames,0,0].mean()
+    focal_y = intrinsic[:num_frames,1,1].mean()
+    focal = (focal_x + focal_y) / 2
+    rend_focal = (rend_intrinsics[0][0,0] + rend_intrinsics[0][1,1]) * 518 / 2
+    focal_scale = rend_focal / focal
+    target_intrinsic = intrinsic[num_frames:].clone()
+    fxy = (target_intrinsic[:,0,0] + target_intrinsic[:,1,1]) / 2 * focal_scale 
+    target_intrinsic[:,0,0] = fxy
+    target_intrinsic[:,1,1] = fxy
+    return target_extrinsic, target_intrinsic
+
+def refine_pose_mast3r(rend_image_pil, target_image_pil, original_size, fxy, target_extrinsic, rend_depth):
+    images_mast3r = load_images_new([rend_image_pil, target_image_pil], size=512, square_ok=True)
+    with torch.no_grad():
+        # [수정] mast3r_model 추론 시 cuda:0 명시 (또는 할당된 device)
+        output = inference([tuple(images_mast3r)], mast3r_model, "cuda:0", batch_size=1, verbose=False)
+    view1, pred1 = output['view1'], output['pred1']
+    view2, pred2 = output['view2'], output['pred2']
+    del output
+    desc1, desc2 = pred1['desc'].squeeze(0).detach(), pred2['desc'].squeeze(0).detach()
+
+    # find 2D-2D matches between the two images
+    matches_im0, matches_im1 = fast_reciprocal_NNs(desc1, desc2, subsample_or_initxy1=8,
+                                                device="cuda:0", dist='dot', block_size=2**13)
+
+    # ignore small border around the edge
+    H0, W0 = view1['true_shape'][0]
+    
+    valid_matches_im0 = (matches_im0[:, 0] >= 3) & (matches_im0[:, 0] < int(W0) - 3) & (
+        matches_im0[:, 1] >= 3) & (matches_im0[:, 1] < int(H0) - 3)
+
+    H1, W1 = view2['true_shape'][0]
+    valid_matches_im1 = (matches_im1[:, 0] >= 3) & (matches_im1[:, 0] < int(W1) - 3) & (
+        matches_im1[:, 1] >= 3) & (matches_im1[:, 1] < int(H1) - 3)
+
+    valid_matches = valid_matches_im0 & valid_matches_im1
+    matches_im0, matches_im1 = matches_im0[valid_matches], matches_im1[valid_matches]
+    scale_x = original_size[1] / W0.item()
+    scale_y = original_size[0] / H0.item()
+    for pixel in matches_im1:
+        pixel[0] *= scale_x
+        pixel[1] *= scale_y
+    for pixel in matches_im0:
+        pixel[0] *= scale_x
+        pixel[1] *= scale_y
+    depth_map = rend_depth[0]
+    fx, fy, cx, cy = fxy.item(), fxy.item(), original_size[1]/2, original_size[0]/2  # Example values for focal lengths and principal point
+    K = np.array([
+        [fx, 0, cx],
+        [0, fy, cy],
+        [0, 0, 1]
+    ])
+    dist_eff = np.array([0,0,0,0], dtype=np.float32)
+    predict_c2w_ini = np.linalg.inv(target_extrinsic[0].cpu().numpy())
+    predict_w2c_ini = target_extrinsic[0].cpu().numpy()
+    initial_rvec, _ = cv2.Rodrigues(predict_c2w_ini[:3,:3].astype(np.float32))
+    initial_tvec = predict_c2w_ini[:3,3].astype(np.float32)
+    K_inv = np.linalg.inv(K)
+    height, width = depth_map.shape
+    x_coords, y_coords = np.meshgrid(np.arange(width), np.arange(height))
+    x_flat = x_coords.flatten()
+    y_flat = y_coords.flatten()
+    depth_flat = depth_map.flatten()
+    x_normalized = (x_flat - K[0, 2]) / K[0, 0]
+    y_normalized = (y_flat - K[1, 2]) / K[1, 1]
+    X_camera = depth_flat * x_normalized
+    Y_camera = depth_flat * y_normalized
+    Z_camera = depth_flat
+    points_camera = np.vstack((X_camera, Y_camera, Z_camera, np.ones_like(X_camera)))
+    points_world = predict_c2w_ini @ points_camera
+    X_world = points_world[0, :]
+    Y_world = points_world[1, :]
+    Z_world = points_world[2, :]
+    points_3D = np.vstack((X_world, Y_world, Z_world))
+    scene_coordinates_gs = points_3D.reshape(3, original_size[0], original_size[1])
+    points_3D_at_pixels = np.zeros((matches_im0.shape[0], 3))
+    for i, (x, y) in enumerate(matches_im0):
+        points_3D_at_pixels[i] = scene_coordinates_gs[:, y, x]
+
+    success, rvec, tvec, inliers = cv2.solvePnPRansac(points_3D_at_pixels.astype(np.float32), matches_im1.astype(np.float32), K, \
+                                                      dist_eff,rvec=initial_rvec,tvec=initial_tvec, useExtrinsicGuess=True, reprojectionError=1.0,\
+                                                      iterationsCount=2000,flags=cv2.SOLVEPNP_EPNP)
+    R = perform_rodrigues_transformation(rvec)
+    trans = -R.T @ np.matrix(tvec)
+    predict_c2w_refine = np.eye(4)
+    predict_c2w_refine[:3,:3] = R.T
+    predict_c2w_refine[:3,3] = trans.reshape(3)
+    target_extrinsic_final = torch.tensor(predict_c2w_refine).inverse().cuda()[None].float()
+    return target_extrinsic_final
+
+def pointcloud_registration(rend_image_pil, target_image_pil, original_size, 
+                            fxy, target_extrinsic, rend_depth, target_pointmap, 
+                            down_pcd, pcd):
+    images_mast3r = load_images_new([rend_image_pil, target_image_pil], size=512, square_ok=True)
+    with torch.no_grad():
+        output = inference([tuple(images_mast3r)], mast3r_model, "cuda:0", batch_size=1, verbose=False)
+    view1, pred1 = output['view1'], output['pred1']
+    view2, pred2 = output['view2'], output['pred2']
+    del output
+    desc1, desc2 = pred1['desc'].squeeze(0).detach(), pred2['desc'].squeeze(0).detach()
+
+    # find 2D-2D matches between the two images
+    matches_im0, matches_im1 = fast_reciprocal_NNs(desc1, desc2, subsample_or_initxy1=8,
+                                                device="cuda:0", dist='dot', block_size=2**13)
+
+    # ignore small border around the edge
+    H0, W0 = view1['true_shape'][0]
+    
+    valid_matches_im0 = (matches_im0[:, 0] >= 3) & (matches_im0[:, 0] < int(W0) - 3) & (
+        matches_im0[:, 1] >= 3) & (matches_im0[:, 1] < int(H0) - 3)
+
+    H1, W1 = view2['true_shape'][0]
+    valid_matches_im1 = (matches_im1[:, 0] >= 3) & (matches_im1[:, 0] < int(W1) - 3) & (
+        matches_im1[:, 1] >= 3) & (matches_im1[:, 1] < int(H1) - 3)
+
+    valid_matches = valid_matches_im0 & valid_matches_im1
+    matches_im0, matches_im1 = matches_im0[valid_matches], matches_im1[valid_matches]
+    scale_x = original_size[1] / W0.item()
+    scale_y = original_size[0] / H0.item()
+    for pixel in matches_im1:
+        pixel[0] *= scale_x
+        pixel[1] *= scale_y
+    for pixel in matches_im0:
+        pixel[0] *= scale_x
+        pixel[1] *= scale_y
+    depth_map = rend_depth[0]
+    fx, fy, cx, cy = fxy.item(), fxy.item(), original_size[1]/2, original_size[0]/2  # Example values for focal lengths and principal point
+    K = np.array([
+        [fx, 0, cx],
+        [0, fy, cy],
+        [0, 0, 1]
+    ])
+    dist_eff = np.array([0,0,0,0], dtype=np.float32)
+    predict_c2w_ini = np.linalg.inv(target_extrinsic[0].cpu().numpy())
+    predict_w2c_ini = target_extrinsic[0].cpu().numpy()
+    initial_rvec, _ = cv2.Rodrigues(predict_c2w_ini[:3,:3].astype(np.float32))
+    initial_tvec = predict_c2w_ini[:3,3].astype(np.float32)
+    K_inv = np.linalg.inv(K)
+    height, width = depth_map.shape
+    x_coords, y_coords = np.meshgrid(np.arange(width), np.arange(height))
+    x_flat = x_coords.flatten()
+    y_flat = y_coords.flatten()
+    depth_flat = depth_map.flatten()
+    x_normalized = (x_flat - K[0, 2]) / K[0, 0]
+    y_normalized = (y_flat - K[1, 2]) / K[1, 1]
+    X_camera = depth_flat * x_normalized
+    Y_camera = depth_flat * y_normalized
+    Z_camera = depth_flat
+    points_camera = np.vstack((X_camera, Y_camera, Z_camera, np.ones_like(X_camera)))
+    points_world = predict_c2w_ini @ points_camera
+    X_world = points_world[0, :]
+    Y_world = points_world[1, :]
+    Z_world = points_world[2, :]
+    points_3D = np.vstack((X_world, Y_world, Z_world))
+    scene_coordinates_gs = points_3D.reshape(3, original_size[0], original_size[1])
+    points_3D_at_pixels = np.zeros((matches_im0.shape[0], 3))
+    for i, (x, y) in enumerate(matches_im0):
+        points_3D_at_pixels[i] = scene_coordinates_gs[:, y, x]
+    
+    points_3D_at_pixels_2 = np.zeros((matches_im1.shape[0], 3))
+    for i, (x, y) in enumerate(matches_im1):
+        points_3D_at_pixels_2[i] = target_pointmap[:, y, x]
+
+    dist_1 = np.linalg.norm(points_3D_at_pixels - points_3D_at_pixels.mean(axis=0), axis=1)
+    scale_1 = dist_1[dist_1 < np.percentile(dist_1, 99)].mean()
+    dist_2 = np.linalg.norm(points_3D_at_pixels_2 - points_3D_at_pixels_2.mean(axis=0), axis=1)
+    scale_2 = dist_2[dist_2 < np.percentile(dist_2, 99)].mean()
+    # scale_1 = np.linalg.norm(points_3D_at_pixels - points_3D_at_pixels.mean(axis=0), axis=1).mean()
+    # scale_2 = np.linalg.norm(points_3D_at_pixels_2 - points_3D_at_pixels_2.mean(axis=0), axis=1).mean()
+    points_3D_at_pixels_2 = points_3D_at_pixels_2 * (scale_1 / scale_2)
+    pcd_1 = o3d.geometry.PointCloud()
+    pcd_1.points = o3d.utility.Vector3dVector(points_3D_at_pixels)
+    pcd_2 = o3d.geometry.PointCloud()
+    pcd_2.points = o3d.utility.Vector3dVector(points_3D_at_pixels_2)
+    indices = np.arange(points_3D_at_pixels.shape[0])
+    correspondences = np.stack([indices, indices], axis=1)
+    correspondences = o3d.utility.Vector2iVector(correspondences)
+    result = o3d.pipelines.registration.registration_ransac_based_on_correspondence(
+        pcd_2,
+        pcd_1,
+        correspondences,
+        0.03,
+        estimation_method=o3d.pipelines.registration.TransformationEstimationPointToPoint(False),
+        ransac_n=5,
+        criteria=o3d.pipelines.registration.RANSACConvergenceCriteria(10000, 10000),
+    )
+    transformation_matrix = result.transformation.copy()
+    transformation_matrix[:3,:3] = transformation_matrix[:3,:3] * (scale_1 / scale_2)
+    evaluation = o3d.pipelines.registration.evaluate_registration(
+        down_pcd, pcd, 0.02, transformation_matrix
+    )
+    return transformation_matrix, evaluation.fitness
+
+# [수정] @spaces.GPU 제거
 def generate_and_extract_glb(
     multiimages: List[Tuple[Image.Image, str]],
     seed: int,
@@ -101,12 +345,23 @@ def generate_and_extract_glb(
     multiimage_algo: Literal["multidiffusion", "stochastic"],
     mesh_simplify: float,
     texture_size: int,
+    refine: Literal["Yes", "No"],
+    ss_refine: Literal["noise", "deltav", "No"],
+    registration_num_frames: int,
+    trellis_stage1_lr: float, 
+    trellis_stage1_start_t: float,  
+    trellis_stage2_lr: float, 
+    trellis_stage2_start_t: float,
     req: gr.Request,
 ) -> Tuple[dict, str, str, str]:
+    """
+    Convert an image to a 3D model and extract GLB file.
+    """
     user_dir = os.path.join(TMP_DIR, str(req.session_hash))
     image_files = [image[0] for image in multiimages]
 
-    outputs, _, _ = pipeline.run(
+    # Generate 3D model
+    outputs, coords, ss_noise = pipeline.run(
         image=image_files,
         seed=seed,
         formats=["gaussian", "mesh"],
@@ -121,24 +376,216 @@ def generate_and_extract_glb(
         },
         mode=multiimage_algo,
     )
+    if refine == "Yes":
+        try:
+            images, alphas = load_and_preprocess_images(multiimages)
+            images, alphas = images.to(device), alphas.to(device)
+            with torch.no_grad():
+                with torch.cuda.amp.autocast(dtype=pipeline.VGGT_dtype):
+                    images = images[None]
+                    # [수정] 분산 배치된 VGGT_model 접근
+                    vggt = pipeline.VGGT_model if not hasattr(pipeline.VGGT_model, 'module') else pipeline.VGGT_model.module
+                    aggregated_tokens_list, ps_idx = vggt.aggregator(images)
+                # Predict Cameras
+                pose_enc = vggt.camera_head(aggregated_tokens_list)[-1]
+                # Extrinsic and intrinsic matrices, following OpenCV convention (camera from world)
+                extrinsic, intrinsic = pose_encoding_to_extri_intri(pose_enc, images.shape[-2:])
+                # Predict Point Cloud
+                point_map, point_conf = vggt.point_head(aggregated_tokens_list, images, ps_idx)
+                del aggregated_tokens_list
+                mask = (alphas[:,0,...][...,None] > 0.8)
+                conf_threshold = np.percentile(point_conf.cpu().numpy(), 50)
+                confidence_mask = (point_conf[0] > conf_threshold) & (point_conf[0] > 1e-5)
+                mask = mask & confidence_mask[...,None]
+                point_map_by_unprojection = point_map[0]
+                point_map_clean = point_map_by_unprojection[mask[...,0]]
+                center_point = point_map_clean.mean(0)
+                scale = np.percentile((point_map_clean - center_point[None]).norm(dim=-1).cpu().numpy(), 98)
+                outlier_mask = (point_map_by_unprojection - center_point[None]).norm(dim=-1) <= scale
+                final_mask = mask & outlier_mask[...,None]
+                point_map_perframe = (point_map_by_unprojection - center_point[None, None, None]) / (2 * scale)
+                point_map_perframe[~final_mask[...,0]] = 127/255
+                point_map_perframe = point_map_perframe.permute(0,3,1,2)
+                images = images[0].permute(0,2,3,1)
+                images[~(alphas[:,0,...][...,None] > 0.8)[...,0]] = 0.
+                input_images = images.permute(0,3,1,2).clone()
+                vggt_extrinsic = extrinsic[0]
+                vggt_extrinsic = torch.cat([vggt_extrinsic, torch.tensor([[[0,0,0,1]]]).repeat(vggt_extrinsic.shape[0], 1, 1).to(vggt_extrinsic)], dim=1)
+                vggt_intrinsic = intrinsic[0]
+                vggt_intrinsic[:,:2] = vggt_intrinsic[:,:2] / 518
+                vggt_extrinsic[:,:3,3] = (torch.matmul(vggt_extrinsic[:,:3,:3], center_point[None,:,None].float())[...,0] + vggt_extrinsic[:,:3,3]) / (2 * scale)
+                pointcloud = point_map_perframe.permute(0,2,3,1)[final_mask[...,0]]
+                idxs = torch.randperm(pointcloud.shape[0])[:min(50000, pointcloud.shape[0])]
+                pcd = o3d.geometry.PointCloud()
+                pcd.points = o3d.utility.Vector3dVector(pointcloud[idxs].cpu().numpy())
+                cl, ind = pcd.remove_statistical_outlier(nb_neighbors=30, std_ratio=3.0)
+                inlier_cloud = pcd.select_by_index(ind)
+                outlier_cloud = pcd.select_by_index(ind, invert=True)
+                distance = np.array(inlier_cloud.points) - np.array(inlier_cloud.points).mean(axis=0)[None]
+                scale = np.percentile(np.linalg.norm(distance, axis=1), 97)
+                voxel_size = 1/64*scale*2
+                down_pcd = inlier_cloud.voxel_down_sample(voxel_size)
+            torch.cuda.empty_cache()
 
+            video, rend_extrinsics, rend_intrinsics = render_utils.render_multiview(outputs['gaussian'][0], num_frames=registration_num_frames)
+            rend_extrinsics = torch.stack(rend_extrinsics, dim=0)
+            rend_intrinsics = torch.stack(rend_intrinsics, dim=0)
+            target_extrinsics = []
+            target_intrinsics = []
+            target_transforms = []
+            target_fitnesses = []   
+            pcd = o3d.geometry.PointCloud()
+            mesh = outputs['mesh'][0]
+            idxs = torch.randperm(mesh.vertices.shape[0])[:min(50000, mesh.vertices.shape[0])]
+            pcd.points = o3d.utility.Vector3dVector(mesh.vertices[idxs].cpu().numpy())
+            distance = np.array(pcd.points) - np.array(pcd.points).mean(axis=0)[None]
+            scale = np.linalg.norm(distance, axis=1).max()
+            voxel_size = 1/64*scale*2
+            pcd = pcd.voxel_down_sample(voxel_size)
+            # pcd.points = o3d.utility.Vector3dVector((coords[:,1:].cpu().numpy() + 0.5) / 64 - 0.5)
+
+            for k in range(len(image_files)):
+                images = torch.stack([TF.ToTensor()(render_image) for render_image in video['color']] + [TF.ToTensor()(image_files[k].convert("RGB"))], dim=0)
+                # if len(images) == 0:
+                with torch.no_grad():
+                    with torch.cuda.amp.autocast(dtype=pipeline.VGGT_dtype):
+                        # predictions = vggt_model(images.cuda())
+                        vggt = pipeline.VGGT_model if not hasattr(pipeline.VGGT_model, 'module') else pipeline.VGGT_model.module
+                        aggregated_tokens_list, ps_idx = vggt.aggregator(images[None].cuda())
+                    pose_enc = vggt.camera_head(aggregated_tokens_list)[-1]
+                extrinsic, intrinsic = pose_encoding_to_extri_intri(pose_enc, images.shape[-2:])
+                extrinsic, intrinsic = extrinsic[0], intrinsic[0]
+                extrinsic = torch.cat([extrinsic, torch.tensor([0,0,0,1])[None,None].repeat(extrinsic.shape[0], 1, 1).to(extrinsic.device)], dim=1)
+                del aggregated_tokens_list, ps_idx
+
+                target_extrinsic, target_intrinsic = align_camera(registration_num_frames, extrinsic, intrinsic, rend_extrinsics, rend_intrinsics)
+                fxy = target_intrinsic[:,0,0]
+                target_intrinsic_tmp = target_intrinsic.clone()
+                target_intrinsic_tmp[:,:2] = target_intrinsic_tmp[:,:2] / 518
+
+                target_extrinsic_list = [target_extrinsic]
+                iou_list = []
+                iterations = 3
+                for i in range(iterations + 1):
+                    j = 0
+                    rend = render_utils.render_frames(outputs['gaussian'][0], target_extrinsic, target_intrinsic_tmp, {'resolution': 518, 'bg_color': (0, 0, 0)}, need_depth=True)
+                    rend_image = rend['color'][j] # (518, 518, 3)
+                    rend_depth = rend['depth'][j] # (3, 518, 518)
+
+                    depth_single = rend_depth[0].astype(np.float32)   # (H, W)
+                    mask = (depth_single != 0).astype(np.uint8)  # 
+                    kernel = np.ones((3, 3), np.uint8)
+                    mask_eroded = cv2.erode(mask, kernel, iterations=3)
+                    depth_eroded = depth_single * mask_eroded
+                    rend_depth_eroded = np.stack([depth_eroded]*3, axis=0)
+
+                    rend_image = torch.tensor(rend_image).permute(2,0,1) / 255
+                    target_image = images[registration_num_frames:].to(target_extrinsic.device)[j]
+                    original_size = (rend_image.shape[1], rend_image.shape[2])
+                    
+                    import torchvision
+                    torchvision.utils.save_image(rend_image, 'rend_image_{}.png'.format(k))
+                    torchvision.utils.save_image(target_image, 'target_image_{}.png'.format(k))
+                    
+                    mask_rend = (rend_image.detach().cpu() > 0).any(dim=0)
+                    mask_target = (target_image.detach().cpu() > 0).any(dim=0)
+                    intersection = (mask_rend & mask_target).sum().item()
+                    union = (mask_rend | mask_target).sum().item()
+                    iou = intersection / union if union > 0 else 0.0
+                    iou_list.append(iou)
+
+                    if i == iterations:
+                        break
+
+                    rend_image = rend_image * torch.from_numpy(mask_eroded[None]).to(rend_image.device)
+                    rend_image_pil = Image.fromarray((rend_image.permute(1,2,0).cpu().numpy() * 255).astype(np.uint8))
+                    target_image_pil = Image.fromarray((target_image.permute(1,2,0).cpu().numpy() * 255).astype(np.uint8))
+                    target_extrinsic[j:j+1] = refine_pose_mast3r(rend_image_pil, target_image_pil, original_size, fxy[j:j+1], target_extrinsic[j:j+1], rend_depth_eroded)    
+                    target_extrinsic_list.append(target_extrinsic[j:j+1])
+                
+                idx = iou_list.index(max(iou_list))
+                target_extrinsic[j:j+1] = target_extrinsic_list[idx]
+                target_transform, fitness = pointcloud_registration(rend_image_pil, target_image_pil, original_size, fxy[j:j+1], target_extrinsic[j:j+1], \
+                                                                    rend_depth_eroded, point_map_perframe[k].cpu().numpy(), down_pcd, pcd)
+                target_transforms.append(target_transform)
+                target_fitnesses.append(fitness)
+                
+                target_extrinsics.append(target_extrinsic[j:j+1])
+                target_intrinsics.append(target_intrinsic_tmp[j:j+1])
+            target_extrinsics = torch.cat(target_extrinsics, dim=0)
+            target_intrinsics = torch.cat(target_intrinsics, dim=0)
+            
+            target_fitnesses_filtered = [x for x in target_fitnesses if x <= 1]
+            idx = target_fitnesses.index(max(target_fitnesses_filtered))
+            target_transform = target_transforms[idx]
+            down_pcd_align = copy.deepcopy(down_pcd).transform(target_transform)
+            # pcd = o3d.geometry.PointCloud()
+            # pcd.points = o3d.utility.Vector3dVector(coords[:,1:].cpu().numpy() / 64 - 0.5)
+            reg_p2p = o3d.pipelines.registration.registration_icp(
+                down_pcd_align, pcd, 0.02, np.eye(4),
+                o3d.pipelines.registration.TransformationEstimationPointToPoint(with_scaling=True),
+                o3d.pipelines.registration.ICPConvergenceCriteria(max_iteration = 10000))
+            down_pcd_align_2 = copy.deepcopy(down_pcd_align).transform(reg_p2p.transformation)
+            input_points = torch.tensor(np.asarray(down_pcd_align_2.points)).to(extrinsic.device).float()
+            input_points = ((input_points + 0.5).clip(0, 1) * 64 - 0.5).to(torch.int32)
+            
+            outputs = pipeline.run_refine(
+                image=image_files,
+                ss_learning_rate=trellis_stage1_lr,
+                ss_start_t=trellis_stage1_start_t,
+                apperance_learning_rate=trellis_stage2_lr,
+                apperance_start_t=trellis_stage2_start_t,
+                extrinsics=target_extrinsics,
+                intrinsics=target_intrinsics,
+                ss_noise=ss_noise,
+                input_points=input_points,
+                ss_refine_type = ss_refine,
+                coords=coords if ss_refine == "No" else None,
+                seed=seed,
+                formats=["mesh", "gaussian"],
+                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:
+            print(f"Error during refinement: {e}")
+    # 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']
     video = [np.concatenate([video[i], video_geo[i]], axis=1) for i in range(len(video))]
     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)
+    
     torch.cuda.empty_cache()
     return state, video_path, glb_path, glb_path
 
+
+# [수정] @spaces.GPU 제거
 def extract_gaussian(state: dict, req: gr.Request) -> Tuple[str, str]:
+    """
+    Extract a Gaussian splatting file from the generated 3D model.
+    """
     user_dir = os.path.join(TMP_DIR, str(req.session_hash))
     gs, _ = unpack_state(state)
     gaussian_path = os.path.join(user_dir, 'sample.ply')
@@ -146,16 +593,15 @@ def extract_gaussian(state: dict, req: gr.Request) -> Tuple[str, str]:
     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:
         _images = []
         for i in range(1, 9):
-            path = f'assets/example_multi_image/{case}_{i}.png'
-            if os.path.exists(path):
-                img = Image.open(path)
+            if os.path.exists(f'assets/example_multi_image/{case}_{i}.png'):
+                img = Image.open(f'assets/example_multi_image/{case}_{i}.png')
                 W, H = img.size
                 img = img.resize((int(W / H * 512), 512))
                 _images.append(np.array(img))
@@ -163,83 +609,195 @@ 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]:
-    image = np.array(image)
-    alpha = image[..., 3]
+    """
+    Split a multi-view image into separate view images.
+    """
+    image_np = np.array(image)
+    # [안정성 추가] 채널 3개짜리(RGB) 이미지가 들어올 경우 에러 방지
+    if image_np.shape[-1] < 4:
+        return [preprocess_image(image)]
+        
+    alpha = image_np[..., 3]
     alpha = np.any(alpha>0, axis=0)
     start_pos = np.where(~alpha[:-1] & alpha[1:])[0].tolist()
     end_pos = np.where(alpha[:-1] & ~alpha[1:])[0].tolist()
     images = []
     for s, e in zip(start_pos, end_pos):
-        images.append(Image.fromarray(image[:, s:e+1]))
+        images.append(Image.fromarray(image_np[:, s:e+1]))
     return [preprocess_image(image) for image in images]
 
-# --- Gradio UI ---
+# Create interface
 demo = gr.Blocks(
     title="ReconViaGen",
-    css=".slider .inner { width: 5px; background: #FFF; } .viewport { aspect-ratio: 4/3; }"
+    css="""
+        .slider .inner { width: 5px; background: #FFF; }
+        .viewport { aspect-ratio: 4/3; }
+        .tabs button.selected { font-size: 20px !important; color: crimson !important; }
+        h1, h2, h3 { text-align: center; display: block; }
+        .md_feedback li { margin-bottom: 0px !important; }
+    """
 )
-
 with demo:
-    gr.Markdown("# 💻 ReconViaGen\n✨This demo is partial. Stay tuned!✨")
+    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():
             with gr.Tabs() as input_tabs:
-                with gr.Tab(label="Input Video or Images", id=0):
+                with gr.Tab(label="Input Video or Images", id=0) as multiimage_input_tab:
                     input_video = gr.Video(label="Upload Video", interactive=True, height=300)
-                    image_prompt = gr.Image(label="Image Prompt", visible=False, type="pil", height=300)
-                    multiimage_prompt = gr.Gallery(label="Image Prompt", columns=3)
-            with gr.Accordion(label="Settings", open=False):
+                    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)
                 randomize_seed = gr.Checkbox(label="Randomize Seed", value=False)
-                ss_guidance_strength = gr.Slider(0.0, 10.0, label="SS Guidance", value=7.5)
-                ss_sampling_steps = gr.Slider(1, 50, label="SS Steps", value=30)
-                slat_guidance_strength = gr.Slider(0.0, 10.0, label="Slat Guidance", value=3.0)
-                slat_sampling_steps = gr.Slider(1, 50, label="Slat Steps", value=12)
-                multiimage_algo = gr.Radio(["stochastic", "multidiffusion"], value="multidiffusion")
-                mesh_simplify = gr.Slider(0.9, 0.98, value=0.95)
-                texture_size = gr.Slider(512, 2048, value=1024, step=512)
+                gr.Markdown("Stage 1: Sparse Structure Generation")
+                with gr.Row():
+                    ss_guidance_strength = gr.Slider(0.0, 10.0, label="Guidance Strength", value=7.5, step=0.1)
+                    ss_sampling_steps = gr.Slider(1, 50, label="Sampling Steps", value=30, step=1)
+                gr.Markdown("Stage 2: Structured Latent Generation")
+                with gr.Row():
+                    slat_guidance_strength = gr.Slider(0.0, 10.0, label="Guidance Strength", value=3.0, step=0.1)
+                    slat_sampling_steps = gr.Slider(1, 50, label="Sampling Steps", value=12, step=1)
+                multiimage_algo = gr.Radio(["stochastic", "multidiffusion"], label="Multi-image Algorithm", value="multidiffusion")
+                refine = gr.Radio(["Yes", "No"], label="Refinement of Not", value="Yes")
+                ss_refine = gr.Radio(["noise", "deltav", "No"], label="Sparse Structure refinement of not", value="No")
+                registration_num_frames = gr.Slider(20, 50, label="Number of frames in registration", value=30, step=1)
+                trellis_stage1_lr = gr.Slider(1e-4, 1., label="trellis_stage1_lr", value=1e-1, step=5e-4)
+                trellis_stage1_start_t = gr.Slider(0., 1., label="trellis_stage1_start_t", value=0.5, step=0.01)
+                trellis_stage2_lr = gr.Slider(1e-4, 1., label="trellis_stage2_lr", value=1e-1, step=5e-4)
+                trellis_stage2_start_t = gr.Slider(0., 1., label="trellis_stage2_start_t", value=0.5, step=0.01)
+
+            with gr.Accordion(label="GLB Extraction Settings", open=False):
+                mesh_simplify = gr.Slider(0.9, 0.98, label="Simplify", value=0.95, step=0.01)
+                texture_size = gr.Slider(512, 2048, label="Texture Size", value=1024, step=512)
+
             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")
-            model_output = LitModel3D(label="Extracted GLB/Gaussian")
+            video_output = gr.Video(label="Generated 3D Asset", autoplay=True, loop=True, height=300)
+            model_output = LitModel3D(label="Extracted GLB/Gaussian", exposure=10.0, height=300)
+            
             with gr.Row():
-                download_glb = gr.DownloadButton("Download GLB", interactive=False)
-                download_gs = gr.DownloadButton("Download Gaussian", interactive=False)
-    output_buf = gr.State()
+                download_glb = gr.DownloadButton(label="Download GLB", interactive=False)
+                download_gs = gr.DownloadButton(label="Download Gaussian", interactive=False)  
     
-    with gr.Row():
-        gr.Examples(examples=prepare_multi_example(), inputs=[image_prompt], fn=split_image, outputs=[multiimage_prompt], run_on_click=True)
+    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(),
+            inputs=[image_prompt],
+            fn=split_image,
+            outputs=[multiimage_prompt],
+            run_on_click=True,
+            examples_per_page=8,
+        )
 
+    # Handlers
     demo.load(start_session)
     demo.unload(end_session)
-    input_video.upload(preprocess_videos, inputs=[input_video], outputs=[multiimage_prompt])
-    input_video.clear(lambda: (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]).then(
+
+    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],
+    ).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],
+        inputs=[multiimage_prompt, seed, ss_guidance_strength, ss_sampling_steps, 
+                slat_guidance_strength, slat_sampling_steps, multiimage_algo, 
+                mesh_simplify, texture_size, refine, ss_refine, registration_num_frames,
+                trellis_stage1_lr, trellis_stage1_start_t, trellis_stage2_lr, 
+                trellis_stage2_start_t],
         outputs=[output_buf, video_output, model_output, download_glb],
-    ).then(lambda: (gr.Button(interactive=True), gr.Button(interactive=True)), outputs=[extract_gs_btn, download_glb])
-    extract_gs_btn.click(extract_gaussian, inputs=[output_buf], outputs=[model_output, download_gs])
+    ).then(
+        lambda: tuple([gr.Button(interactive=True), gr.Button(interactive=True)]),
+        outputs=[extract_gs_btn, download_glb],
+    )
 
-# --- 메인 실행부 (VRAM 분산 최적화) ---
+    video_output.clear(
+        lambda: tuple([gr.Button(interactive=False), gr.Button(interactive=False), gr.Button(interactive=False)]),
+        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],
+    )
+
+    model_output.clear(
+        lambda: tuple([gr.Button(interactive=False), gr.Button(interactive=False)]),
+        outputs=[download_glb, download_gs],
+    )
+    
+
+# Launch the Gradio app - VRAM 4개 분산 최적화 적용
 if __name__ == "__main__":
-    # 모델 로드
     pipeline = TrellisVGGTTo3DPipeline.from_pretrained("Stable-X/trellis-vggt-v0-2")
     
     num_gpus = torch.cuda.device_count()
     if num_gpus >= 4:
-        print(f"--- 4 GPUs Detected: Splitting Models to prevent VRAM Error ---")
-        # 모델의 각 부분을 서로 다른 GPU 메모리에 적재하여 1개 GPU의 부담을 줄임
-        pipeline.to("cuda:0") 
-        if hasattr(pipeline, 'VGGT_model'): pipeline.VGGT_model.to("cuda:1")
-        if hasattr(pipeline, 'birefnet_model'): pipeline.birefnet_model.to("cuda:2")
-        # 가장 무거운 디코더들은 3번 GPU로 격리
-        if hasattr(pipeline, 'slat_decoder'): pipeline.slat_decoder.to("cuda:3")
-        if hasattr(pipeline, 'sparse_structure_decoder'): pipeline.sparse_structure_decoder.to("cuda:3")
+        # [VRAM 분산 핵심] 각 모델을 물리적으로 다른 GPU에 로드하여 OOM 방지
+        pipeline.to("cuda:0") # 메인 파이프라인
+        if hasattr(pipeline, 'VGGT_model'): 
+            pipeline.VGGT_model.to("cuda:1")
+        if hasattr(pipeline, 'birefnet_model'): 
+            pipeline.birefnet_model.to("cuda:2")
+        # 가장 무거운 디코더들을 3번 GPU로 격리
+        if hasattr(pipeline, 'slat_decoder'): 
+            pipeline.slat_decoder.to("cuda:3")
+        if hasattr(pipeline, 'sparse_structure_decoder'): 
+            pipeline.sparse_structure_decoder.to("cuda:3")
+            
+        # Mast3r 모델 로드 (cuda:0 사용)
+        mast3r_model = AsymmetricMASt3R.from_pretrained("naver/MASt3R_ViTLarge_BaseDecoder_512_catmlpdpt_metric").to("cuda:0").eval()
+        print("Success: 4 GPU VRAM Sharding Activated.")
     else:
         pipeline.cuda()
-
+        mast3r_model = AsymmetricMASt3R.from_pretrained("naver/MASt3R_ViTLarge_BaseDecoder_512_catmlpdpt_metric").cuda().eval()
+        
     demo.launch()