Documentation Index
Fetch the complete documentation index at: https://mintlify.com/oktopuzSlid/detectorPlacas/llms.txt
Use this file to discover all available pages before exploring further.
DetectorPlacas provides three independent detection modes — static images, video files, and live webcam — each implemented in its own Python script. Every mode loads the same frozen_inference_graph.pb and labelmap.pbtxt, but each has a distinct confidence threshold tuned for its use case: stricter for images where precision matters most, and progressively more permissive for video and real-time webcam capture where recall is more valuable.
Script: Object_detection_image.pyThe image script accepts a --ruta flag pointing to a directory of .jpg files. When the flag is omitted or empty, the script falls back to the current working directory. It collects every file ending in .jpg using os.listdir() combined with endswith('.jpg'), then runs inference on each one in sequence. A new window opens for each image; press any key to advance to the next.| Setting | Value |
|---|
| Flag | --ruta (path to image directory) |
| File filter | os.listdir() + .endswith('.jpg') |
| Confidence threshold | min_score_thresh=0.65 |
| Line thickness | 8 |
| Window title | 'Detector de Placas PRUEBA-IMAGEN' |
| Advance | Press any key |
# Detect plates in a specific directory
python Object_detection_image.py --ruta=/path/to/images
# Detect plates in the current working directory
python Object_detection_image.py
image = cv2.imread(os.path.join(FLAGS.ruta, a))
image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
image_expanded = np.expand_dims(image_rgb, axis=0)
(boxes, scores, classes, num) = sess.run(
[detection_boxes, detection_scores, detection_classes, num_detections],
feed_dict={image_tensor: image_expanded})
vis_util.visualize_boxes_and_labels_on_image_array(
image,
np.squeeze(boxes),
np.squeeze(classes).astype(np.int32),
np.squeeze(scores),
category_index,
use_normalized_coordinates=True,
line_thickness=8,
min_score_thresh=0.65)
cv2.imshow('Detector de Placas PRUEBA-IMAGEN', image)
cv2.waitKeyEx()
Script: Object_detection_video.pyThe video script opens a file named prueba.mp4 from the project root (VIDEO_NAME = 'prueba.mp4'). It reads frames in a loop with video.read(), runs inference on each frame, draws bounding boxes, and displays the annotated frame. Press any key to advance to the next frame. When the video ends, resources are released and all windows are closed.| Setting | Value |
|---|
| Input file | prueba.mp4 (hardcoded, must be in project root) |
| Confidence threshold | min_score_thresh=0.55 |
| Line thickness | 5 |
| Window title | 'Detector de placas- PRUEBA VIDEO' |
| Advance | Press any key |
# Rename your video file to prueba.mp4 and place it in the project root
python Object_detection_video.py
video = cv2.VideoCapture(PATH_TO_VIDEO)
while(video.isOpened()):
ret, frame = video.read()
frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
frame_expanded = np.expand_dims(frame_rgb, axis=0)
(boxes, scores, classes, num) = sess.run(
[detection_boxes, detection_scores, detection_classes, num_detections],
feed_dict={image_tensor: frame_expanded})
vis_util.visualize_boxes_and_labels_on_image_array(
frame,
np.squeeze(boxes),
np.squeeze(classes).astype(np.int32),
np.squeeze(scores),
category_index,
use_normalized_coordinates=True,
line_thickness=5,
min_score_thresh=0.55)
cv2.imshow('Detector de placas- PRUEBA VIDEO', frame)
cv2.waitKeyEx()
video.release()
cv2.destroyAllWindows()
Script: Object_detection_webcam.pyThe webcam script opens camera index 0 — your system’s default webcam — and sets the capture resolution to 1280×720. It runs in an infinite loop: each iteration reads a frame, converts it from BGR to RGB, runs inference, draws detections, and displays the result. Press any key to advance to the next frame.| Setting | Value |
|---|
| Camera index | 0 (cv2.VideoCapture(0)) |
| Resolution | 1280×720 |
| Confidence threshold | min_score_thresh=0.50 |
| Line thickness | 8 |
| Window title | 'Detector de placas PRUEBA-WEBCAM' |
| Advance | Press any key |
python Object_detection_webcam.py
video = cv2.VideoCapture(0)
ret = video.set(3, 1280)
ret = video.set(4, 720)
while(True):
ret, frame = video.read()
frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
frame_expanded = np.expand_dims(frame_rgb, axis=0)
(boxes, scores, classes, num) = sess.run(
[detection_boxes, detection_scores, detection_classes, num_detections],
feed_dict={image_tensor: frame_expanded})
vis_util.visualize_boxes_and_labels_on_image_array(
frame,
np.squeeze(boxes),
np.squeeze(classes).astype(np.int32),
np.squeeze(scores),
category_index,
use_normalized_coordinates=True,
line_thickness=8,
min_score_thresh=0.50)
cv2.imshow('Detector de placas PRUEBA-WEBCAM', frame)
cv2.waitKeyEx()
Each detection mode uses a different confidence threshold, intentionally tuned for its context:| Mode | Threshold | Rationale |
|---|
| Image | 0.65 | Highest precision — reduces false positives on static images |
| Video | 0.55 | Balanced — trades some precision for better recall across frames |
| Webcam | 0.50 | Most permissive — maximises recall for real-time detection |
All three modes convert each frame from OpenCV’s native BGR color order to RGB with cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) before creating the input tensor. This conversion is required because the TensorFlow model expects RGB input, while cv2.imread and cv2.VideoCapture both return BGR arrays by default.