I am wondering how to merge these two code bases, so I can dynamically change the ROI, while still using the degirum SDK for inference. Tried a few solutions did not land on anything: regards!
<opencv_code>
import cv2
import datetime
cap = cv2.VideoCapture(1)
if not cap.isOpened():
print(“Camera cannot open”)
exit()
cv2.namedWindow(“Camera”, cv2.WND_PROP_FULLSCREEN)
cv2.setWindowProperty(“Camera”, cv2.WND_PROP_FULLSCREEN, cv2.WINDOW_FULLSCREEN)
SCREEN_W, SCREEN_H = 1280, 720
globalz = {
‘zoom_step’ : 0.01,
‘zoom_xy_min’ : 0.0,
‘zoom_x’ : 0.0,
‘zoom_y’ : 0.0,
‘zoom_xy_max’ : 0.4,
‘zoom_wh_min’ : 1.0,
‘zoom_wh’ : 1.0,
‘zoom_wh_max’ : 0.2,
}
overlay_visible = True
annotation_text = “”
annotation_timer = 0
def set_annotation(text, duration=60):
global annotation_text, annotation_timer
annotation_text = text
annotation_timer = duration
def apply_roi(frame):
h, w = frame.shape[:2]
x = int(globalz[‘zoom_x’] * w)
y = int(globalz[‘zoom_y’] * h)
cw = int(globalz[‘zoom_wh’] * w)
ch = int(globalz[‘zoom_wh’] * h)
x = max(0, min(x, w - 1))
y = max(0, min(y, h - 1))
cw = max(1, min(cw, w - x))
ch = max(1, min(ch, h - y))
cropped = frame[y:y+ch, x:x+cw]
return cv2.resize(cropped, (SCREEN_W, SCREEN_H))
def zoom_in():
if (globalz[‘zoom_x’] + globalz[‘zoom_step’] > globalz[‘zoom_xy_max’] and
globalz[‘zoom_y’] + globalz[‘zoom_step’] > globalz[‘zoom_xy_max’]):
set_annotation(“At max zoom”)
else:
globalz[‘zoom_x’] += globalz[‘zoom_step’]
globalz[‘zoom_y’] += globalz[‘zoom_step’]
globalz[‘zoom_wh’] -= globalz[‘zoom_step’] * 2
set_annotation(“Zoom In”)
def zoom_out():
if (globalz[‘zoom_x’] - globalz[‘zoom_step’] < globalz[‘zoom_xy_min’] and
globalz[‘zoom_y’] - globalz[‘zoom_step’] < globalz[‘zoom_xy_min’]):
set_annotation(“At min zoom”)
else:
globalz[‘zoom_x’] -= globalz[‘zoom_step’]
globalz[‘zoom_y’] -= globalz[‘zoom_step’]
globalz[‘zoom_wh’] += globalz[‘zoom_step’] * 2
set_annotation(“Zoom Out”)
def set_min_zoom():
globalz[‘zoom_x’] = globalz[‘zoom_xy_min’]
globalz[‘zoom_y’] = globalz[‘zoom_xy_min’]
globalz[‘zoom_wh’] = globalz[‘zoom_wh_min’]
set_annotation(“Zoom reset”)
def pan_right():
globalz[‘zoom_y’] = min(globalz[‘zoom_y’] + globalz[‘zoom_step’], globalz[‘zoom_xy_max’])
set_annotation(“Pan Right”)
def pan_left():
globalz[‘zoom_y’] = max(globalz[‘zoom_y’] - globalz[‘zoom_step’], globalz[‘zoom_xy_min’])
set_annotation(“Pan Left”)
def pan_up():
globalz[‘zoom_x’] = max(globalz[‘zoom_x’] - globalz[‘zoom_step’], globalz[‘zoom_xy_min’])
set_annotation(“Pan Up”)
def pan_down():
globalz[‘zoom_x’] = min(globalz[‘zoom_x’] + globalz[‘zoom_step’], globalz[‘zoom_xy_max’])
set_annotation(“Pan Down”)
def take_picture(frame):
filename = datetime.datetime.now().strftime(“%Y-%m-%d_%H.%M.%S.jpg”)
cv2.imwrite(filename, frame)
set_annotation(f"Saved: {filename}“)
print(f"Picture saved: {filename}”)
def draw_overlay(frame):
global annotation_text, annotation_timer
font = cv2.FONT_HERSHEY_PLAIN
lines = [
“PiGlassV2”,
“P: Take Picture”,
“+/-: Zoom In/Out”,
“Arrow keys: Pan”,
“R: Reset Zoom”,
“O: Toggle Overlay”,
“Q: Quit”,
]
y = SCREEN_H - len(lines) * 30 - 10
for line in lines:
cv2.putText(frame, line, (10, y), font, 2, (255, 255, 255), 2)
y += 30
if annotation_timer > 0:
cv2.putText(frame, annotation_text,
(SCREEN_W // 2 - 200, SCREEN_H // 2),
font, 3, (255, 255, 0), 3)
annotation_timer -= 1
KEY_UP = [82, 2490368, 65362]
KEY_DOWN = [84, 2621440, 65364]
KEY_LEFT = [81, 2424832, 65361]
KEY_RIGHT = [83, 2555904, 65363]
while True:
ret, frame = cap.read()
if not ret:
break
display = apply_roi(frame)
# ← your DeGirum inference goes here, operating on `display`
if overlay_visible:
draw_overlay(display)
cv2.imshow("Camera", display)
key = cv2.waitKey(1) & 0xFFFFFF
if key == ord('q'): break
elif key == ord('+') or key == ord('='): zoom_in()
elif key == ord('-'): zoom_out()
elif key in KEY_UP: pan_up()
elif key in KEY_DOWN: pan_down()
elif key in KEY_LEFT: pan_left()
elif key in KEY_RIGHT: pan_right()
elif key == ord('r'): set_min_zoom()
elif key == ord('p'): take_picture(display)
elif key == ord('o'):
overlay_visible = not overlay_visible
set_annotation("Overlay " + ("ON" if overlay_visible else "OFF"))
cap.release()
cv2.destroyAllWindows() </opencv_code>
<degirum_code>
with degirum_tools.Display(‘AI Camera’) as output_display, degirum_tools.open_video_writer(‘output_license_plate_video.mp4’,
w = 640, h = 480, fps = 30.0) as video_writer:
for detected_license_plates in degirum_tools.predict_stream(lp_det_model, video_source):
if detected_license_plates.results:
cropped_license_plates = crop_images(detected_license_plates.image, detected_license_plates.results)
print(f’cropped license plates: {cropped_license_plates}‘)
for index, cropped_license_plate in enumerate(cropped_license_plates):
ocr_results = lpocr_model.predict(cropped_license_plate)
print(f’MOST IMPORTANT: OCR RESULTS: {ocr_results}’)
license_plate = rearrange_detections(ocr_results.results)
print(f’final OCR text: {license_plate}')
#detected_license_plates.results[index][‘label’] = license_plate
video_writer.write(detected_license_plates.image)
output_display.show(detected_license_plates)
</degirum_code>
