File size: 9,761 Bytes
505117b | 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 | """Create a side-by-side raw and Mask R-CNN segmentation showcase video."""
from __future__ import annotations
import argparse
import hashlib
import json
from pathlib import Path
import cv2
import numpy as np
import torch
from detectron2 import model_zoo
from detectron2.config import get_cfg
from detectron2.engine import DefaultPredictor
CASE_SPECS = (
("5gs_22C", "45", "5 g/s | inlet subcooling 22 deg C | 45 V"),
("10gs_22C", "45V", "10 g/s | inlet subcooling 22 deg C | 45 V"),
("15gs_20C", "45", "15 g/s | inlet subcooling 20 deg C | 45 V"),
("25gs_20C", "45", "25 g/s | inlet subcooling 20 deg C | 45 V"),
)
ROI_X = 0
ROI_Y = 485
ROI_WIDTH = 1024
ROI_HEIGHT = 70
DISPLAY_Y0 = 360
DISPLAY_Y1 = 680
MINIMAL_DISPLAY_Y0 = 430
MINIMAL_DISPLAY_Y1 = 590
FRAME_SIZE = (1920, 720)
PANEL_SIZE = (880, 275)
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--dataset-root", required=True, type=Path)
parser.add_argument("--weights", required=True, type=Path)
parser.add_argument("--output", required=True, type=Path)
parser.add_argument("--frames-per-case", type=int, default=20)
parser.add_argument("--fps", type=float, default=5.0)
parser.add_argument("--minimal-stacked", action="store_true")
parser.add_argument("--score-threshold", type=float, default=0.30)
parser.add_argument("--detections-per-image", type=int, default=300)
parser.add_argument("--device", default="cuda" if torch.cuda.is_available() else "cpu")
return parser.parse_args()
def build_predictor(args: argparse.Namespace) -> DefaultPredictor:
cfg = get_cfg()
cfg.merge_from_file(
model_zoo.get_config_file("COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml")
)
cfg.MODEL.WEIGHTS = str(args.weights)
cfg.MODEL.DEVICE = args.device
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 1
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = args.score_threshold
cfg.MODEL.ANCHOR_GENERATOR.SIZES = [[8], [16], [32], [64], [128]]
cfg.INPUT.MIN_SIZE_TEST = 640
cfg.INPUT.MAX_SIZE_TEST = 900
cfg.TEST.DETECTIONS_PER_IMAGE = args.detections_per_image
return DefaultPredictor(cfg)
def select_frames(image_dir: Path, count: int) -> list[Path]:
images = sorted(image_dir.glob("*.bmp"))
if len(images) < count:
raise ValueError(f"{image_dir} has {len(images)} BMP files; need {count}.")
indices = np.linspace(0, len(images) - 1, count, dtype=int)
return [images[index] for index in indices]
def union_mask(predictor: DefaultPredictor, roi: np.ndarray) -> np.ndarray:
outputs = predictor(roi)
masks = outputs["instances"].pred_masks.to("cpu").numpy()
if len(masks) == 0:
return np.zeros(roi.shape[:2], dtype=np.uint8)
return np.any(masks, axis=0).astype(np.uint8) * 255
def make_overlay(roi: np.ndarray, mask: np.ndarray) -> np.ndarray:
overlay = roi.copy()
red = np.zeros_like(roi)
red[:, :, 2] = 255
selected = mask.astype(bool)
overlay[selected] = cv2.addWeighted(roi, 0.48, red, 0.52, 0)[selected]
contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
cv2.drawContours(overlay, contours, -1, (70, 235, 110), 1)
return overlay
def draw_text(image: np.ndarray, text: str, position: tuple[int, int], scale: float, color: tuple[int, int, int]) -> None:
cv2.putText(image, text, position, cv2.FONT_HERSHEY_SIMPLEX, scale, color, 2, cv2.LINE_AA)
def compose_frame(
raw: np.ndarray,
overlay: np.ndarray,
case_text: str,
progress: str,
display_fps: float,
) -> np.ndarray:
canvas = np.full((FRAME_SIZE[1], FRAME_SIZE[0], 3), (20, 24, 30), dtype=np.uint8)
draw_text(canvas, "BubbleID-Flow: Raw Optical Image and Mask R-CNN Segmentation", (55, 55), 0.92, (245, 245, 245))
draw_text(canvas, case_text, (55, 105), 0.70, (185, 215, 235))
draw_text(canvas, progress, (1420, 105), 0.55, (190, 190, 190))
raw_panel = cv2.resize(raw, PANEL_SIZE, interpolation=cv2.INTER_AREA)
overlay_panel = cv2.resize(overlay, PANEL_SIZE, interpolation=cv2.INTER_AREA)
left_x, right_x, top_y = 70, 970, 180
canvas[top_y : top_y + PANEL_SIZE[1], left_x : left_x + PANEL_SIZE[0]] = raw_panel
canvas[top_y : top_y + PANEL_SIZE[1], right_x : right_x + PANEL_SIZE[0]] = overlay_panel
cv2.rectangle(canvas, (left_x, top_y), (left_x + PANEL_SIZE[0], top_y + PANEL_SIZE[1]), (220, 220, 220), 2)
cv2.rectangle(canvas, (right_x, top_y), (right_x + PANEL_SIZE[0], top_y + PANEL_SIZE[1]), (220, 220, 220), 2)
draw_text(canvas, "RAW OPTICAL IMAGE", (left_x, top_y - 20), 0.66, (245, 245, 245))
draw_text(canvas, "UNION-MASK OVERLAY", (right_x, top_y - 20), 0.66, (245, 245, 245))
draw_text(canvas, "Red: predicted vapor region | Green: union-mask boundary", (55, 540), 0.62, (230, 155, 155))
draw_text(canvas, f"Display uses uniformly sampled frames at {display_fps:g} fps; source acquisition rate: 3000 frames/s.", (55, 595), 0.52, (190, 190, 190))
draw_text(canvas, "Metric: image-plane vapor area fraction, not volumetric void fraction.", (55, 635), 0.52, (190, 190, 190))
return canvas
def compose_minimal_stacked(raw: np.ndarray, overlay: np.ndarray) -> np.ndarray:
canvas = np.zeros((FRAME_SIZE[1], FRAME_SIZE[0], 3), dtype=np.uint8)
panel_size = (1600, 250)
raw_panel = cv2.resize(raw, panel_size, interpolation=cv2.INTER_AREA)
overlay_panel = cv2.resize(overlay, panel_size, interpolation=cv2.INTER_AREA)
left_x, top_y, bottom_y = 160, 70, 400
canvas[top_y : top_y + panel_size[1], left_x : left_x + panel_size[0]] = raw_panel
canvas[bottom_y : bottom_y + panel_size[1], left_x : left_x + panel_size[0]] = overlay_panel
draw_text(canvas, "RAW", (left_x, 45), 0.70, (245, 245, 245))
draw_text(canvas, "SEGMENTED", (left_x, 375), 0.70, (245, 245, 245))
return canvas
def sha256(path: Path) -> str:
digest = hashlib.sha256()
with path.open("rb") as stream:
for chunk in iter(lambda: stream.read(1024 * 1024), b""):
digest.update(chunk)
return digest.hexdigest()
def main() -> None:
args = parse_args()
if args.frames_per_case <= 0 or args.fps <= 0:
raise ValueError("frames-per-case and fps must be positive")
image_root = args.dataset_root / "MultimodalData" / "Images"
if not image_root.is_dir():
raise FileNotFoundError(f"Expected image root: {image_root}")
if not args.weights.is_file():
raise FileNotFoundError(f"Weights not found: {args.weights}")
args.output.parent.mkdir(parents=True, exist_ok=True)
predictor = build_predictor(args)
fourcc = cv2.VideoWriter_fourcc(*"mp4v")
writer = cv2.VideoWriter(str(args.output), fourcc, args.fps, FRAME_SIZE)
if not writer.isOpened():
raise RuntimeError(f"Could not open video writer for {args.output}")
selected: dict[str, list[str]] = {}
try:
total_frames = len(CASE_SPECS) * args.frames_per_case
frame_number = 0
for case_id, state_id, label in CASE_SPECS:
source_files = select_frames(image_root / case_id / state_id, args.frames_per_case)
selected[case_id] = [path.name for path in source_files]
for image_path in source_files:
image = cv2.imread(str(image_path), cv2.IMREAD_COLOR)
if image is None or image.shape[:2] != (1024, 1024):
raise ValueError(f"Expected a 1024 x 1024 readable image: {image_path}")
roi = image[ROI_Y : ROI_Y + ROI_HEIGHT, ROI_X : ROI_X + ROI_WIDTH]
mask = union_mask(predictor, roi)
if args.minimal_stacked:
display_y0, display_y1 = MINIMAL_DISPLAY_Y0, MINIMAL_DISPLAY_Y1
else:
display_y0, display_y1 = DISPLAY_Y0, DISPLAY_Y1
display_raw = image[display_y0:display_y1, :]
display_overlay = display_raw.copy()
y0 = ROI_Y - display_y0
display_overlay[y0 : y0 + ROI_HEIGHT, :] = make_overlay(roi, mask)
frame_number += 1
if args.minimal_stacked:
writer.write(compose_minimal_stacked(display_raw, display_overlay))
else:
writer.write(
compose_frame(
display_raw,
display_overlay,
label,
f"Frame {frame_number}/{total_frames}",
args.fps,
)
)
finally:
writer.release()
manifest = {
"dataset_root": str(args.dataset_root),
"weights": str(args.weights),
"weights_sha256": sha256(args.weights),
"output": str(args.output),
"duration_seconds": len(CASE_SPECS) * args.frames_per_case / args.fps,
"fps": args.fps,
"frames_per_case": args.frames_per_case,
"score_threshold": args.score_threshold,
"detections_per_image": args.detections_per_image,
"roi_xywh": [ROI_X, ROI_Y, ROI_WIDTH, ROI_HEIGHT],
"display_crop_y": [
MINIMAL_DISPLAY_Y0 if args.minimal_stacked else DISPLAY_Y0,
MINIMAL_DISPLAY_Y1 if args.minimal_stacked else DISPLAY_Y1,
],
"minimal_stacked": args.minimal_stacked,
"case_order": [spec[0] for spec in CASE_SPECS],
"selected_frames": selected,
}
manifest_path = args.output.with_suffix(".json")
manifest_path.write_text(json.dumps(manifest, indent=2), encoding="utf-8")
print(f"Wrote {args.output}")
print(f"Wrote {manifest_path}")
if __name__ == "__main__":
main()
|