Spaces:
Sleeping
Sleeping
File size: 12,465 Bytes
a6beb58 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 |
import cv2
import numpy as np
import torch
from segment_anything import sam_model_registry, SamPredictor
from ultralytics import YOLO
import mediapipe as mp
import json
from datetime import datetime
import time
import os
def process_seat_depth_analysis(image_path, eye_to_ear_cm=7.0, sam_checkpoint="sam_vit_b_01ec64.pth"):
"""
Main function to process seat depth analysis
Args:
image_path: Path to the input image
eye_to_ear_cm: Real-world eye to ear distance for scaling (default 7.0 cm)
sam_checkpoint: Path to SAM model checkpoint
Returns:
tuple: (output_json, pose_image, seat_band_image, final_image)
"""
start_time = time.time()
def put_text_safe(image, text, org, font, font_scale, color, thickness):
text_size, _ = cv2.getTextSize(text, font, font_scale, thickness)
text_width, text_height = text_size
x, y = org
h, w = image.shape[:2]
# Adjust X if text goes out on the right
if x + text_width > w:
x = w - text_width - 5 # 5 pixel padding from right
# Adjust X if text goes out on the left
if x < 0:
x = 5 # 5 pixel padding from left
# Adjust Y if text goes above image
if y - text_height < 0:
y = text_height + 5 # push down
# Adjust Y if text goes below image
if y > h:
y = h - 5
cv2.putText(image, text, (x, y), font, font_scale, color, thickness)
# === Load image ===
image_bgr = cv2.imread(image_path)
if image_bgr is None:
raise ValueError(f"Could not load image from {image_path}")
image_rgb = cv2.cvtColor(image_bgr, cv2.COLOR_BGR2RGB)
h, w = image_rgb.shape[:2]
# === Run MediaPipe Pose Detection ===
mp_pose = mp.solutions.pose
pose = mp_pose.Pose(static_image_mode=True)
results = pose.process(image_rgb)
if not results.pose_landmarks:
raise ValueError("No pose detected in the image")
landmarks = results.pose_landmarks.landmark
# === Get Knee and Eye X,Y coordinates ===
left_knee = landmarks[25]
right_knee = landmarks[26]
left_eye = landmarks[2]
right_eye = landmarks[5]
right_ear = landmarks[8]
left_ear = landmarks[7]
left_hip = landmarks[23]
right_hip = landmarks[24]
# Convert to pixel coordinates
left_knee_px = (int(left_knee.x * w), int(left_knee.y * h))
right_knee_px = (int(right_knee.x * w), int(right_knee.y * h))
left_eye_px = (int(left_eye.x * w), int(left_eye.y * h))
right_eye_px = (int(right_eye.x * w), int(right_eye.y * h))
left_ear_px = (int(left_ear.x * w), int(left_ear.y * h))
right_ear_px = (int(right_ear.x * w), int(right_ear.y * h))
left_hip_px = (int(left_hip.x * w), int(left_hip.y * h))
right_hip_px = (int(right_hip.x * w), int(right_hip.y * h))
# === Determine Facing Direction ===
avg_knee_x = (left_knee_px[0] + right_knee_px[0]) / 2
avg_eye_x = (left_eye_px[0] + right_eye_px[0]) / 2
facing_direction = "right" if avg_knee_x > avg_eye_x else "left"
# === Create Pose Overlay (Image 1) ===
pose_image = image_rgb.copy()
mp_drawing = mp.solutions.drawing_utils
mp_drawing_styles = mp.solutions.drawing_styles
mp_drawing.draw_landmarks(
pose_image,
results.pose_landmarks,
mp_pose.POSE_CONNECTIONS,
landmark_drawing_spec=mp_drawing_styles.get_default_pose_landmarks_style()
)
# === Step 1: Detect Chair with YOLOv8 ===
yolo_model = YOLO("yolov8n.pt")
yolo_results = yolo_model(image_rgb)
# === Step 2: Get Chair Box ===
chair_box = None
chair_confidence = 0.0
for result in yolo_results:
for box, cls, conf in zip(result.boxes.xyxy, result.boxes.cls, result.boxes.conf):
if int(cls.item()) == 56: # 56 = chair
chair_box = box.cpu().numpy().astype(int)
chair_confidence = float(conf.item())
break
if chair_box is None:
raise ValueError("No chair detected in the image")
x1, y1, x2, y2 = chair_box
chair_height = y2 - y1
adjusted_y1 = y1 + int(0.25 * chair_height)
input_box = np.array([x1, adjusted_y1, x2, y2])
# === Step 3: Load SAM ===
model_type = "vit_b"
device = "cuda" if torch.cuda.is_available() else "cpu"
sam = sam_model_registry[model_type](checkpoint=sam_checkpoint)
sam.to(device=device)
predictor = SamPredictor(sam)
# === Step 4: Predict Mask from Bounding Box ===
predictor.set_image(image_rgb)
masks, scores, _ = predictor.predict(box=input_box[None, :], multimask_output=True)
best_mask = masks[np.argmax(scores)]
# === Step 5: Largest Component Only ===
def get_largest_connected_component(mask):
num_labels, labels, stats, _ = cv2.connectedComponentsWithStats(mask.astype(np.uint8), connectivity=8)
if num_labels <= 1:
return mask
largest_label = 1 + np.argmax(stats[1:, cv2.CC_STAT_AREA])
return labels == largest_label
cleaned_mask = get_largest_connected_component(best_mask)
# === Step 6: Estimate Seat Front ===
knee_y = int((left_knee_px[1] + right_knee_px[1]) / 2)
band_thickness = chair_height // 2
y_min = max(0, knee_y - band_thickness)
y_max = min(h, knee_y + band_thickness)
band = cleaned_mask[y_min:y_max, :]
chair_pixels_x = np.where(band)[1]
if chair_pixels_x.size == 0:
raise ValueError("No chair pixels detected at knee level")
seat_front_x = chair_pixels_x.max() if facing_direction == "right" else chair_pixels_x.min()
seat_front_y = knee_y
# === Create Seat Front Band Visualization (Image 2) ===
seat_band_image = image_rgb.copy()
cv2.line(seat_band_image, (0, y_min), (w, y_min), (0, 255, 0), 2)
cv2.line(seat_band_image, (0, y_max), (w, y_max), (0, 255, 0), 2)
cv2.circle(seat_band_image, (seat_front_x, seat_front_y), 8, (0, 0, 255), -1)
put_text_safe(seat_band_image, "Seat Front", (seat_front_x + 10, seat_front_y - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
# === Calculate Back of Knee Position ===
def euclidean_distance(p1, p2):
return np.linalg.norm(np.array(p1) - np.array(p2))
# Calculate thigh length for proportional offset
if facing_direction == "right":
hip_pt = right_hip_px
knee_pt_original = right_knee_px
else:
hip_pt = left_hip_px
knee_pt_original = left_knee_px
thigh_length_px = euclidean_distance(hip_pt, knee_pt_original)
# Back of knee is typically 12-15% of thigh length behind knee center
back_of_knee_offset = thigh_length_px * 0.13 # 13% of thigh length
# Apply offset in the backward direction
if facing_direction == "right":
knee_pt = (int(knee_pt_original[0] - back_of_knee_offset), knee_pt_original[1])
else:
knee_pt = (int(knee_pt_original[0] + back_of_knee_offset), knee_pt_original[1])
# === Calculate Measurements ===
clearance_px = abs(seat_front_x - knee_pt[0])
# Check visibility and calculate eye-to-ear distance
visibility_warnings = []
if facing_direction == "right" and (right_eye.visibility < 0.5 or right_ear.visibility < 0.5):
visibility_warnings.append("Right eye or ear not clearly visible. Scaling may be inaccurate.")
elif facing_direction == "left" and (left_eye.visibility < 0.5 or left_ear.visibility < 0.5):
visibility_warnings.append("Left eye or ear not clearly visible. Scaling may be inaccurate.")
if facing_direction == "right":
eye_coord = right_eye_px
ear_coord = right_ear_px
else:
eye_coord = left_eye_px
ear_coord = left_ear_px
eye_to_ear_px = euclidean_distance(eye_coord, ear_coord)
pixels_per_cm = eye_to_ear_px / eye_to_ear_cm
clearance_cm = clearance_px / pixels_per_cm
# Determine if back of knee is behind seat front
if facing_direction == "right":
knee_behind_seat = knee_pt[0] < seat_front_x
else:
knee_behind_seat = knee_pt[0] > seat_front_x
# === Classification ===
category = "Too Short"
if knee_behind_seat or clearance_cm < 2:
if clearance_cm < 2:
category = "Too Deep"
reasoning = f"Clearance of {clearance_cm:.2f}cm is less than 2cm minimum"
elif knee_behind_seat:
category = "Too Deep"
reasoning = "Back of knee is behind seat front"
elif clearance_cm <= 6:
category = "Optimal"
reasoning = f"Clearance of {clearance_cm:.2f}cm falls within optimal range (2-6cm)"
else:
category = "Too Short"
reasoning = f"Clearance of {clearance_cm:.2f}cm exceeds 6cm optimal maximum"
# === Create Final Visualization (Image 3) ===
final_image = image_rgb.copy()
# Draw seat front and knee
cv2.circle(final_image, (seat_front_x, seat_front_y), 8, (0, 0, 255), -1)
cv2.circle(final_image, knee_pt, 8, (255, 0, 0), -1)
# Height at which the line floats
line_y = min(seat_front_y, knee_pt[1]) - 30
# Draw horizontal line (floating)
cv2.line(final_image, (min(seat_front_x, knee_pt[0]), line_y),
(max(seat_front_x, knee_pt[0]), line_y),
(255, 255, 0), 2)
# Add arrow tips
cv2.arrowedLine(final_image,
(min(seat_front_x, knee_pt[0]) + 20, line_y),
(min(seat_front_x, knee_pt[0]), line_y),
(255, 255, 0), 2, tipLength=0.4)
cv2.arrowedLine(final_image,
(max(seat_front_x, knee_pt[0]) - 20, line_y),
(max(seat_front_x, knee_pt[0]), line_y),
(255, 255, 0), 2, tipLength=0.4)
# Put clearance text above the line
put_text_safe(final_image, f"Knee clearance: {clearance_cm:.1f} cm",
(min(seat_front_x, knee_pt[0]) + 10, line_y - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 255), 2)
# Draw eye-to-ear line
cv2.line(final_image, eye_coord, ear_coord, (0, 255, 0), 2)
put_text_safe(final_image, f"{eye_to_ear_cm:.1f}cm",
(eye_coord[0], eye_coord[1] - 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 0), 2)
# === Generate JSON Output ===
processing_time = int((time.time() - start_time) * 1000)
output_json = {
"frame_id": os.path.basename(image_path),
"timestamp": datetime.now().isoformat(),
"pose_detection": {
"pose_detected": True,
"facing_direction": facing_direction,
"landmarks_visibility": {
"left_eye": float(left_eye.visibility),
"right_eye": float(right_eye.visibility),
"left_ear": float(left_ear.visibility),
"right_ear": float(right_ear.visibility),
"left_knee": float(left_knee.visibility),
"right_knee": float(right_knee.visibility),
"left_hip": float(left_hip.visibility),
"right_hip": float(right_hip.visibility)
}
},
"chair_detection": {
"chair_detected": True,
"chair_bbox": chair_box.tolist(),
"chair_confidence": chair_confidence
},
"measurements": {
"eye_to_ear_distance_px": float(eye_to_ear_px),
"eye_to_ear_distance_cm": float(eye_to_ear_cm),
"pixels_per_cm": float(pixels_per_cm),
"seat_front_position": [int(seat_front_x), int(seat_front_y)],
"back_of_knee_position": [int(knee_pt[0]), int(knee_pt[1])],
"knee_clearance_px": float(clearance_px),
"knee_clearance_cm": float(clearance_cm),
"thigh_length_px": float(thigh_length_px),
"back_of_knee_offset_applied": float(back_of_knee_offset)
},
"classification": {
"category": category,
"knee_behind_seat": bool(knee_behind_seat),
"reasoning": reasoning
},
"debug_info": {
"band_y_range": [int(y_min), int(y_max)],
"chair_pixels_detected": int(chair_pixels_x.size),
"segmentation_success": True,
"scaling_method": "eye_to_ear_reference"
},
"warnings": visibility_warnings,
"processing_time_ms": processing_time
}
return output_json, pose_image, seat_band_image, final_image |