floppyrj45/cosma-qc
2
0
Fork 0
Code Issues 2 Pull Requests Actions Packages Projects Releases Wiki Activity

fix: viser_ply filtrage outliers statistiques — supprime gros pâtés bruités

Browse Source
This commit is contained in:
Flag
2026-04-23 23:10:11 +00:00
parent fb38ff2192
commit 352af149fd
1 changed files with 21 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
scripts/viser_ply.py
View File

@@ -1,15 +1,7 @@
#!/usr/bin/env python3 #!/usr/bin/env python3
"""Minimal standalone viser viewer for a single PLY point cloud. """Minimal standalone viser viewer for a single PLY point cloud."""
Usage:
python3 viser_ply.py path/to/cloud.ply --port 8200
"""
from __future__ import annotations from __future__ import annotations
import argparse, sys, time
import argparse
import sys
import time
import numpy as np import numpy as np
@@ -18,10 +10,9 @@ def main():
ap.add_argument("ply", help="PLY file to visualize") ap.add_argument("ply", help="PLY file to visualize")
ap.add_argument("--port", type=int, default=8200) ap.add_argument("--port", type=int, default=8200)
ap.add_argument("--point-size", type=float, default=0.01) ap.add_argument("--point-size", type=float, default=0.01)
ap.add_argument("--downsample", type=float, default=0.0, ap.add_argument("--downsample", type=float, default=0.0)
help="Voxel downsample size (0 = no downsample)") ap.add_argument("--max-points", type=int, default=2_000_000)
ap.add_argument("--max-points", type=int, default=2_000_000, ap.add_argument("--no-filter", action="store_true", help="Disable outlier removal")
help="Random subsample to this many points if cloud is larger")
args = ap.parse_args() args = ap.parse_args()
try: try:
@@ -31,15 +22,30 @@ def main():
sys.exit(f"missing dep: {e}") sys.exit(f"missing dep: {e}")
pcd = o3d.io.read_point_cloud(args.ply) pcd = o3d.io.read_point_cloud(args.ply)
n_raw = len(pcd.points)
print(f"loaded {n_raw:,} pts from {args.ply}", flush=True)
if args.downsample > 0: if args.downsample > 0:
pcd = pcd.voxel_down_sample(args.downsample) pcd = pcd.voxel_down_sample(args.downsample)
print(f"after voxel downsample: {len(pcd.points):,} pts", flush=True)
if not args.no_filter and len(pcd.points) > 100:
# Remove statistical outliers (blobs / noise from low-confidence frames)
pcd, _ = pcd.remove_statistical_outlier(nb_neighbors=20, std_ratio=2.0)
print(f"after outlier removal: {len(pcd.points):,} pts", flush=True)
# Second pass tighter for dense clouds
if len(pcd.points) > 200_000:
pcd, _ = pcd.remove_statistical_outlier(nb_neighbors=30, std_ratio=1.5)
print(f"after 2nd pass: {len(pcd.points):,} pts", flush=True)
pts = np.asarray(pcd.points, dtype=np.float32) pts = np.asarray(pcd.points, dtype=np.float32)
cols = np.asarray(pcd.colors, dtype=np.float32) if pcd.has_colors() else np.ones_like(pts) * 0.7 cols = np.asarray(pcd.colors, dtype=np.float32) if pcd.has_colors() else np.ones_like(pts) * 0.7
if len(pts) > args.max_points: if len(pts) > args.max_points:
idx = np.random.choice(len(pts), args.max_points, replace=False) idx = np.random.choice(len(pts), args.max_points, replace=False)
pts, cols = pts[idx], cols[idx] pts, cols = pts[idx], cols[idx]
print(f"loaded {len(pts):,} pts from {args.ply}", flush=True)
print(f"displaying {len(pts):,} pts", flush=True)
server = viser.ViserServer(host="0.0.0.0", port=args.port) server = viser.ViserServer(host="0.0.0.0", port=args.port)
server.scene.add_point_cloud( server.scene.add_point_cloud(
"/cloud", points=pts, colors=(cols * 255).astype(np.uint8), point_size=args.point_size "/cloud", points=pts, colors=(cols * 255).astype(np.uint8), point_size=args.point_size