8f05a7c188
- Update ASR, face, OCR, pose processors - Add release pre-flight check script - Add synonym generation, chunk processing scripts - Add face recognition, stamp search utilities
104 lines
3.4 KiB
Python
104 lines
3.4 KiB
Python
#!/opt/homebrew/bin/python3.11
|
|
"""
|
|
Find ANY Small Rectangular Object in Hands
|
|
"""
|
|
|
|
import cv2
|
|
import numpy as np
|
|
import os
|
|
|
|
UUID = "384b0ff44aaaa1f1"
|
|
BASE_DIR = f"output/{UUID}/florence2_results"
|
|
|
|
# Frames to check
|
|
FRAMES = [
|
|
"scan_6756.jpg", # 112:36
|
|
"scan_6763.jpg", # 112:43
|
|
"scan_6790.jpg", # 113:10
|
|
"scan_6813.jpg", # 113:33
|
|
"scan_6832.jpg", # 113:52
|
|
]
|
|
|
|
print("🖐️ Searching for SMALL OBJECTS in hands...")
|
|
|
|
for frame_name in FRAMES:
|
|
img_path = os.path.join(BASE_DIR, frame_name)
|
|
if not os.path.exists(img_path):
|
|
continue
|
|
|
|
img = cv2.imread(img_path)
|
|
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
|
|
|
|
# 1. Hand Detection (Skin Tone) - Adjusted for lighting
|
|
# Broad range to catch hands in shadow or bright light
|
|
skin_mask = cv2.inRange(hsv, np.array([0, 15, 40]), np.array([25, 160, 255]))
|
|
|
|
# Morphological cleaning
|
|
kernel = np.ones((5, 5), np.uint8)
|
|
skin_mask = cv2.morphologyEx(skin_mask, cv2.MORPH_CLOSE, kernel)
|
|
skin_mask = cv2.morphologyEx(skin_mask, cv2.MORPH_OPEN, kernel)
|
|
|
|
# 2. Find contours inside/near hands
|
|
# We dilate the mask slightly to include objects held IN the hand
|
|
skin_dilated = cv2.dilate(skin_mask, kernel, iterations=3)
|
|
|
|
# Find contours in the full image
|
|
contours, _ = cv2.findContours(
|
|
skin_dilated, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE
|
|
)
|
|
|
|
print(f"\n🎞️ Scanning {frame_name}...")
|
|
found_count = 0
|
|
|
|
for cnt in contours:
|
|
x, y, w, h = cv2.boundingRect(cnt)
|
|
area = cv2.contourArea(cnt)
|
|
|
|
# Object size filter:
|
|
# Too small (< 100px) = noise
|
|
# Too big (> 15000px) = likely the face or body part itself
|
|
if 100 < area < 15000:
|
|
# Shape filter: Rectangle-like (Aspect ratio 0.5 to 2.0)
|
|
aspect_ratio = float(w) / h
|
|
|
|
# Check for rectangularity (Extent)
|
|
rect_area = w * h
|
|
if rect_area > 0:
|
|
extent = float(area) / rect_area
|
|
# If extent > 0.4, it's somewhat rectangular/filled
|
|
if 0.5 < aspect_ratio < 2.5 and extent > 0.4:
|
|
found_count += 1
|
|
print(
|
|
f" ✅ Candidate Object: Area={int(area)}, Pos=({x},{y}), Size={w}x{h}"
|
|
)
|
|
|
|
# Crop with padding
|
|
pad = 10
|
|
crop = img[
|
|
max(0, y - pad) : min(img.shape[0], y + h + pad),
|
|
max(0, x - pad) : min(img.shape[1], x + w + pad),
|
|
]
|
|
crop_path = os.path.join(
|
|
BASE_DIR, f"object_in_hand_{frame_name}_{x}_{y}.jpg"
|
|
)
|
|
cv2.imwrite(crop_path, crop)
|
|
|
|
# Draw on main image
|
|
cv2.rectangle(img, (x, y), (x + w, y + h), (0, 255, 0), 3)
|
|
cv2.putText(
|
|
img,
|
|
f"OBJ? ({int(area)})",
|
|
(x, y - 10),
|
|
cv2.FONT_HERSHEY_SIMPLEX,
|
|
0.6,
|
|
(0, 255, 0),
|
|
2,
|
|
)
|
|
|
|
if found_count == 0:
|
|
print(" ❌ No small objects found in hands.")
|
|
else:
|
|
res_path = os.path.join(BASE_DIR, f"result_objects_{frame_name}")
|
|
cv2.imwrite(res_path, img)
|
|
print(f" 🎨 Result saved to {res_path}")
|