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 ships four Python scripts: three detection scripts that each handle a different input source (static images, a video file, and a live webcam feed) and one Tkinter GUI launcher that wraps all three. This page documents every flag, constant, visualization setting, and behavioral detail drawn directly from the source code for each script.
Object_detection_image.py
Object_detection_video.py
Object_detection_webcam.py
interfaz.py
Detects license plates in all .jpg files found in a target directory. The script iterates over every matching file, runs the frozen-graph inference pipeline on each image in sequence, and opens an OpenCV window to display the annotated result before advancing to the next file.Run commandpython Object_detection_image.py --ruta=/path/to/images
tf.app.flags:flags = tf.app.flags
flags.DEFINE_string('ruta', '', 'Path to the CSV input')
FLAGS = flags.FLAGS
Path to the directory containing .jpg image files. When left empty the script falls back to os.getcwd(), so detection runs against the current working directory.
| Name | Value | Purpose |
|---|
MODEL_NAME | 'inference_graph' | Subdirectory containing the frozen graph |
NUM_CLASSES | 1 | Only one class: license plate |
PATH_TO_CKPT | os.path.join(CWD_PATH, MODEL_NAME, 'frozen_inference_graph.pb') | Full path to the frozen .pb model file |
PATH_TO_LABELS | os.path.join(CWD_PATH, 'labelmap.pbtxt') | Full path to the label map |
PATH_TO_IMAGE | os.path.join(CWD_PATH) | Base directory for image discovery |
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)
| Setting | Value |
|---|
min_score_thresh | 0.65 — detections below 65 % confidence are suppressed |
line_thickness | 8 px bounding-box border |
use_normalized_coordinates | True — boxes are in [0, 1] range |
'Detector de Placas PRUEBA-IMAGEN'Behavior: Uses os.listdir() to collect every .jpg file under --ruta. For each image, runs a sess.run() inference pass, annotates the result with vis_util, then calls cv2.waitKeyEx() — press any key to advance to the next image. cv2.destroyAllWindows() is called after all images have been processed. Detects license plates in a video file frame by frame. The script opens the video with cv2.VideoCapture, processes each frame through the same frozen-graph pipeline, and displays the annotated frame in an OpenCV window.Run commandpython Object_detection_video.py
| Name | Value | Purpose |
|---|
MODEL_NAME | 'inference_graph' | Subdirectory containing the frozen graph |
VIDEO_NAME | 'prueba.mp4' | Expected in the current working directory |
NUM_CLASSES | 1 | Only one class: license plate |
PATH_TO_CKPT | os.path.join(CWD_PATH, MODEL_NAME, 'frozen_inference_graph.pb') | Full path to the frozen .pb model file |
PATH_TO_LABELS | os.path.join(CWD_PATH, 'labelmap.pbtxt') | Full path to the label map |
PATH_TO_VIDEO | os.path.join(CWD_PATH, VIDEO_NAME) | Full path to prueba.mp4 |
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)
| Setting | Value |
|---|
min_score_thresh | 0.55 — lower threshold than image mode to account for motion blur |
line_thickness | 5 px bounding-box border |
use_normalized_coordinates | True |
'Detector de placas- PRUEBA VIDEO'Behavior: Opens prueba.mp4 via cv2.VideoCapture(PATH_TO_VIDEO) and iterates with while(video.isOpened()). Each iteration reads a frame with video.read(), converts BGR→RGB, runs inference, annotates, and calls cv2.waitKeyEx(). When the video ends, the loop exits, video.release() frees the capture handle, and cv2.destroyAllWindows() closes all windows. Detects license plates in real time from the system webcam. The script captures frames in a continuous loop, running inference on each frame and overlaying detection results live.Run commandpython Object_detection_webcam.py
| Name | Value | Purpose |
|---|
MODEL_NAME | 'inference_graph' | Subdirectory containing the frozen graph |
NUM_CLASSES | 1 | Only one class: license plate |
PATH_TO_CKPT | os.path.join(CWD_PATH, MODEL_NAME, 'frozen_inference_graph.pb') | Full path to the frozen .pb model file |
PATH_TO_LABELS | os.path.join(CWD_PATH, 'labelmap.pbtxt') | Full path to the label map |
video = cv2.VideoCapture(0) # index 0 = default system webcam
ret = video.set(3, 1280) # width → 1280 px
ret = video.set(4, 720) # height → 720 px
| Property | Value |
|---|
| Camera index | 0 (default system webcam) |
| Capture width | 1280 px |
| Capture height | 720 px |
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)
| Setting | Value |
|---|
min_score_thresh | 0.50 — widest threshold across all three scripts |
line_thickness | 8 px bounding-box border |
use_normalized_coordinates | True |
'Detector de placas PRUEBA-WEBCAM'Behavior: Runs in an infinite while(True) loop with no break condition. Each iteration captures a frame, converts BGR→RGB, runs a sess.run() inference pass, annotates the frame, and calls cv2.waitKeyEx(). Cleanup lines (video.release() and cv2.destroyAllWindows()) are present in the source after the loop but are unreachable at runtime because the loop has no exit path. To stop the script, press Ctrl+C in the terminal. A Tkinter GUI launcher that wraps all three detection scripts. Running interfaz.py opens a small desktop window with radio buttons to select the detection mode, an optional path entry for image mode, and a single button to launch the selected script via os.system().Run commandWindow title: "Proyecto"Module-level side effect: Before the GUI enters its event loop, the script executes os.system('python ejemplo.py') at module level. This call runs unconditionally whenever interfaz.py is imported or executed directly.os.system('python ejemplo.py')
| Element | Details |
|---|
| Header label | Text: "Detección de Placas", font=courier, bg=azure |
selected variable | IntVar() — holds the active radio selection |
| Radio button — IMAGEN | value=1 |
| Radio button — VIDEO | value=2 |
| Radio button — CAMARA | value=3 |
| RUTA entry | entry2 — Entry widget; path supplied here is forwarded to --ruta in image mode |
| PRUEBA button | Calls clicked() on press |
clicked() dispatch functiondef clicked():
path = str("python Object_detection_image.py --ruta={}".format(entry2.get()))
if (selected.get()==1):
print(path)
os.system(path)
if (selected.get()==2):
os.system('python Object_detection_video.py')
if (selected.get()==3):
os.system('python Object_detection_webcam.py')
selected == 1, the value from the RUTA entry (entry2) is interpolated into the --ruta flag and the image detection script is launched. Selections 2 and 3 invoke the video and webcam scripts respectively, ignoring the RUTA entry. Shared TensorFlow tensors
All three detection scripts retrieve the same five tensors from the loaded detection_graph using get_tensor_by_name(). These names are fixed by the TF Object Detection API export format.
| Tensor name | Role | Shape / notes |
|---|
image_tensor:0 | Input — the image batch fed to the model | [batch, H, W, 3] — RGB uint8 values |
detection_boxes:0 | Output — predicted bounding boxes | [batch, num_detections, 4] — normalized [ymin, xmin, ymax, xmax] in range [0, 1] |
detection_scores:0 | Output — confidence score per detection | [batch, num_detections] — float in [0, 1] |
detection_classes:0 | Output — class index per detection | [batch, num_detections] — 1 = license plate; cast to int32 with .astype(np.int32) |
num_detections:0 | Output — count of valid detections returned | [batch] — integer scalar per image in the batch |
image_tensor = detection_graph.get_tensor_by_name('image_tensor:0')
detection_boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
detection_scores = detection_graph.get_tensor_by_name('detection_scores:0')
detection_classes = detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = detection_graph.get_tensor_by_name('num_detections:0')