File size: 12,524 Bytes
7c29f6d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 | """
Visualization utilities for inference
"""
import os
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.animation as animation
from matplotlib.gridspec import GridSpec
from mpl_toolkits.mplot3d import Axes3D
from mpl_toolkits.mplot3d.art3d import Poly3DCollection
from pathlib import Path
try:
import bpy
from mathutils import Vector, Matrix, Euler
BLENDER_AVAILABLE = True
except ImportError:
BLENDER_AVAILABLE = False
def visualize_input_data(input_data, save_path="input_data_vis.png"):
"""
Visualize input data including colors, normals (as colors), and ref_pcd.
Args:
input_data: Dictionary containing ref_shape_pcd, ref_shape_rgbs, ref_shape_normals,
ref_pcd, ref_rgb, ref_normal
save_path: Path to save the visualization image
"""
fig = plt.figure(figsize=(14, 10))
gs = GridSpec(2, 2, figure=fig)
ax1 = fig.add_subplot(gs[0, 0], projection='3d')
points1 = input_data['ref_shape_pcd'][0].detach().cpu().numpy()
rgbs1 = input_data['ref_shape_rgbs'][0].detach().cpu().numpy()
rgbs1_clip = np.clip(rgbs1, 0, 1)
ax1.scatter(points1[:, 0], points1[:, 1], points1[:, 2], c=rgbs1_clip, s=1)
ax1.set_title("ref_shape_pcd RGB")
ax1.set_xlabel('X')
ax1.set_ylabel('Y')
ax1.set_zlabel('Z')
ax1.set_xlim([-0.5, 0.5])
ax1.set_ylim([-0.5, 0.5])
ax1.set_zlim([-0.5, 0.5])
ax2 = fig.add_subplot(gs[0, 1], projection='3d')
normals1 = input_data['ref_shape_normals'][0].detach().cpu().numpy()
normals_vis1 = (normals1 + 1) * 0.5
ax2.scatter(points1[:, 0], points1[:, 1], points1[:, 2], c=normals_vis1, s=1)
ax2.set_title("ref_shape_pcd Normals (as color)")
ax2.set_xlabel('X')
ax2.set_ylabel('Y')
ax2.set_zlabel('Z')
ax2.set_xlim([-0.5, 0.5])
ax2.set_ylim([-0.5, 0.5])
ax2.set_zlim([-0.5, 0.5])
ax3 = fig.add_subplot(gs[1, 0], projection='3d')
points2 = input_data['ref_pcd'][0].detach().cpu().numpy()
rgbs2 = input_data['ref_rgb'][0].detach().cpu().numpy()
rgbs2_clip = np.clip(rgbs2, 0, 1)
ax3.scatter(points2[:, 0], points2[:, 1], points2[:, 2], c=rgbs2_clip, s=3)
ax3.set_title("ref_pcd RGB")
ax3.set_xlabel('X')
ax3.set_ylabel('Y')
ax3.set_zlabel('Z')
ax3.set_xlim([-0.5, 0.5])
ax3.set_ylim([-0.5, 0.5])
ax3.set_zlim([-0.5, 0.5])
ax4 = fig.add_subplot(gs[1, 1], projection='3d')
normals2 = input_data['ref_normal'][0].detach().cpu().numpy()
normals_vis2 = (normals2 + 1) * 0.5
ax4.scatter(points2[:, 0], points2[:, 1], points2[:, 2], c=normals_vis2, s=3)
ax4.set_title("ref_pcd Normals (as color)")
ax4.set_xlabel('X')
ax4.set_ylabel('Y')
ax4.set_zlabel('Z')
ax4.set_xlim([-0.5, 0.5])
ax4.set_ylim([-0.5, 0.5])
ax4.set_zlim([-0.5, 0.5])
plt.tight_layout()
plt.savefig(save_path, dpi=180)
plt.close(fig)
print(f"Input data visualization saved to: {save_path}")
def visualize_pointcloud_prediction(result, sample, save_dir="./visualization_results", idx = 0, fps=8):
"""
Visualize point cloud prediction results.
Args:
result: Model prediction results containing input_data, rgb, pcd_moved, loss_metrics.
sample: Input data containing obj_name, rgb_video, point_clouds, uv_gt, color_gt.
save_dir: Directory to save visualization results.
fps: Animation frame rate.
"""
save_dir = Path(save_dir)
save_dir.mkdir(exist_ok=True, parents=True)
gt_point_clouds = sample['point_clouds'][idx].float().cpu().numpy() # Shape: (T, N_points, D_features)
pred_point_clouds = result['pcd_moved'][idx].float().cpu().numpy() # Shape: (T, N_points, 3)
rgb_frames = sample['rgb_video'][idx].float().cpu().numpy() # Shape: (T, H, W, C)
obj_name = sample['obj_name'][idx]
T, N_points, _ = gt_point_clouds.shape
if gt_point_clouds.shape[-1] > 3:
gt_coords = gt_point_clouds[..., :3]
else:
gt_coords = gt_point_clouds
pred_coords = pred_point_clouds
all_coords = np.concatenate([gt_coords.reshape(-1, 3), pred_coords.reshape(-1, 3)], axis=0)
min_vals = np.min(all_coords, axis=0)
max_vals = np.max(all_coords, axis=0)
center = (min_vals + max_vals) / 2
max_range = np.max(max_vals - min_vals) / 2 * 1.2
fig = plt.figure(figsize=(24, 6))
gs = GridSpec(1, 4, figure=fig)
ax1 = fig.add_subplot(gs[0, 0], projection='3d')
ax1.set_title('Ground Truth', fontsize=14, fontweight='bold')
ax2 = fig.add_subplot(gs[0, 1], projection='3d')
ax2.set_title('Prediction', fontsize=14, fontweight='bold')
ax3 = fig.add_subplot(gs[0, 2], projection='3d')
ax3.set_title('GT (Blue) vs Pred (Red)', fontsize=14, fontweight='bold')
ax4 = fig.add_subplot(gs[0, 3])
ax4.set_title('RGB Video', fontsize=14, fontweight='bold')
ax4.axis('off')
for ax in [ax1, ax2, ax3]:
ax.set_xlim([center[0] - max_range, center[0] + max_range])
ax.set_ylim([center[1] - max_range, center[1] + max_range])
ax.set_zlim([center[2] - max_range, center[2] + max_range])
ax.set_xlabel('X')
ax.set_ylabel('Y')
ax.set_zlabel('Z')
if len(rgb_frames.shape) == 4 and rgb_frames.shape[-1] == 3:
if rgb_frames.max() > 1.0:
rgb_frames_normalized = rgb_frames / 255.0
else:
rgb_frames_normalized = rgb_frames
else:
print(f"Warning: RGB frames shape {rgb_frames.shape} is unexpected")
rgb_frames_normalized = np.zeros((T, 224, 224, 3))
im = ax4.imshow(rgb_frames_normalized[0])
def animate(frame):
ax1.clear()
ax2.clear()
ax3.clear()
ax1.set_title(f'GT (Frame {frame+1}/{T})', fontsize=14, fontweight='bold')
ax2.set_title(f'Pred (Frame {frame+1}/{T})', fontsize=14, fontweight='bold')
ax3.set_title(f'GT vs Pred (Frame {frame+1}/{T})', fontsize=14, fontweight='bold')
ax4.set_title(f'RGB (Frame {frame+1}/{T})', fontsize=14, fontweight='bold')
for ax in [ax1, ax2, ax3]:
ax.set_xlim([center[0] - max_range, center[0] + max_range])
ax.set_ylim([center[1] - max_range, center[1] + max_range])
ax.set_zlim([center[2] - max_range, center[2] + max_range])
ax.set_xlabel('X')
ax.set_ylabel('Y')
ax.set_zlabel('Z')
gt_coords_frame = gt_coords[frame]
pred_coords_frame = pred_coords[frame]
# GT subplot
ax1.scatter(gt_coords_frame[:, 0], gt_coords_frame[:, 1], gt_coords_frame[:, 2],
c='blue', s=2, alpha=0.6)
# Prediction subplot
ax2.scatter(pred_coords_frame[:, 0], pred_coords_frame[:, 1], pred_coords_frame[:, 2],
c='red', s=2, alpha=0.6)
# Comparison subplot
ax3.scatter(gt_coords_frame[:, 0], gt_coords_frame[:, 1], gt_coords_frame[:, 2],
c='blue', s=2, alpha=0.6, label='GT')
ax3.scatter(pred_coords_frame[:, 0], pred_coords_frame[:, 1], pred_coords_frame[:, 2],
c='red', s=2, alpha=0.6, label='Pred')
ax3.legend()
im.set_array(rgb_frames_normalized[frame])
return []
anim = animation.FuncAnimation(fig, animate, frames=T, interval=1000//fps, blit=False)
save_path = save_dir / f"{obj_name}_combined_visualization.gif"
anim.save(str(save_path), writer='pillow', fps=fps, dpi=100)
plt.close(fig)
mse_per_frame = np.mean((gt_coords - pred_coords) ** 2, axis=(1, 2))
avg_mse = np.mean(mse_per_frame)
return {
'visualization_path': str(save_path),
}
def visualize_point_cloud_motion(ref_pcd, pcd_moved, trajs, save_path):
"""
Visualize point cloud positions before and after movement with motion trajectory arrows.
Args:
ref_pcd: Original point cloud, shape (N, 3)
pcd_moved: Moved point cloud, shape (N, 3)
trajs: Motion trajectory vectors, shape (N, 3)
save_path: Path to save the image
"""
ref_pcd_np = ref_pcd.cpu().numpy()
pcd_moved_np = pcd_moved.cpu().numpy()
trajs_np = trajs.cpu().numpy()
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.scatter(ref_pcd_np[:, 0], ref_pcd_np[:, 1], ref_pcd_np[:, 2], s=5, c='b', label='ref_pcd')
ax.scatter(pcd_moved_np[:, 0], pcd_moved_np[:, 1], pcd_moved_np[:, 2], s=5, c='r', label='pcd_moved')
ax.quiver(
ref_pcd_np[:, 0], ref_pcd_np[:, 1], ref_pcd_np[:, 2],
trajs_np[:, 0], trajs_np[:, 1], trajs_np[:, 2],
length=1.0, normalize=False, color='g', linewidth=0.5, alpha=0.7
)
ax.set_title("Point Cloud Motion Comparison")
ax.legend()
plt.savefig(save_path, bbox_inches='tight')
plt.close(fig)
def plot_smoothing_comparison(displacements_before, displacements_after,
motion_threshold, method, visualization_dir):
"""
Plot comparison of trajectory smoothing before and after.
Args:
displacements_before: List of displacement arrays before smoothing
displacements_after: List of displacement arrays after smoothing
motion_threshold: Motion threshold used for filtering
method: Smoothing method name
visualization_dir: Directory to save the plot
"""
fig, axes = plt.subplots(2, 2, figsize=(14, 10))
point_mean_before = np.mean(np.stack(displacements_before, axis=0), axis=0)
point_mean_after = np.mean(np.stack(displacements_after, axis=0), axis=0)
axes[0, 0].scatter(range(len(point_mean_before)), point_mean_before, s=1, alpha=0.5, color='b', label='Before')
axes[0, 0].scatter(range(len(point_mean_after)), point_mean_after, s=1, alpha=0.5, color='r', label='After')
axes[0, 0].set_xlabel('Point Index')
axes[0, 0].set_ylabel('Mean Displacement Across Frames')
axes[0, 0].set_title('Per-Point Average Displacement')
axes[0, 0].grid(True)
axes[0, 0].legend()
all_before = np.concatenate(displacements_before)
axes[0, 1].hist(all_before, bins=100, color='b', alpha=0.7, edgecolor='black')
axes[0, 1].set_xlabel('Displacement Magnitude')
axes[0, 1].set_ylabel('Frequency')
axes[0, 1].set_title('Displacement Distribution (Before)')
axes[0, 1].axvline(x=motion_threshold, color='r', linestyle='--', label=f'Threshold={motion_threshold}')
axes[0, 1].set_yscale('log')
axes[0, 1].legend()
axes[0, 1].grid(True)
all_after = np.concatenate(displacements_after)
axes[1, 0].hist(all_after, bins=100, color='r', alpha=0.7, edgecolor='black')
axes[1, 0].set_xlabel('Displacement Magnitude')
axes[1, 0].set_ylabel('Frequency')
axes[1, 0].set_title('Displacement Distribution (After)')
axes[1, 0].axvline(x=motion_threshold, color='r', linestyle='--', label=f'Threshold={motion_threshold}')
axes[1, 0].set_yscale('log')
axes[1, 0].legend()
axes[1, 0].grid(True)
point_std_before = np.std(np.stack(displacements_before, axis=0), axis=0)
point_std_after = np.std(np.stack(displacements_after, axis=0), axis=0)
axes[1, 1].scatter(range(len(point_std_before)), point_std_before, s=1, alpha=0.5, color='b', label='Before')
axes[1, 1].scatter(range(len(point_std_after)), point_std_after, s=1, alpha=0.5, color='r', label='After')
axes[1, 1].set_xlabel('Point Index')
axes[1, 1].set_ylabel('Std Displacement Across Frames')
axes[1, 1].set_title('Per-Point Displacement Std')
axes[1, 1].grid(True)
axes[1, 1].legend()
plt.tight_layout()
smooth_plot_path = os.path.join(visualization_dir, f'smooth_comparison_{method}.png')
plt.savefig(smooth_plot_path, dpi=150)
plt.close()
print(f"Smoothing comparison plot saved to: {smooth_plot_path}")
print(f"\n=== Smoothing Statistics ===")
print(f"Average displacement before: {np.mean(all_before):.6f} ± {np.std(all_before):.6f}")
print(f"Average displacement after: {np.mean(all_after):.6f} ± {np.std(all_after):.6f}")
print(f"Jitter reduction rate: {(1 - np.std(all_after) / np.std(all_before)) * 100:.2f}%")
print(f"Per-point average displacement - before: {np.mean(point_mean_before):.6f} ± {np.std(point_mean_before):.6f}")
print(f"Per-point average displacement - after: {np.mean(point_mean_after):.6f} ± {np.std(point_mean_after):.6f}")
|