This project aim to build an on premise licence plate recognition model.
├── LICENSE
├── Makefile <- Makefile with commands like `make requirements`
├── README.md <- The top-level README for developers using this project.
├── data
│ ├── external <- Data from third party sources.
│ ├── interim <- Intermediate data that has been transformed.
│ ├── processed <- The final, canonical data sets for modeling.
│ └── raw <- The original, immutable data dump.
│
├── docs <- A default Sphinx project; see sphinx-doc.org for details
│
├── models <- scripts to predict and train on models
│
├── notebooks <- Jupyter notebooks. Naming convention is a number (for ordering),
│ the creator's initials, and a short `-` delimited description, e.g.
│ `1.0-jqp-initial-data-exploration`.
│
├── reports <- Generated analysis as HTML, PDF, LaTeX, etc.
│ └── figures <- Generated graphics and figures to be used in reporting
│
├── requirements.txt <- The requirements file for reproducing the analysis environment, e.g.
│ generated with `pip freeze > requirements.txt`
│
├── setup.py <- makes project pip installable (pip install -e .) so src can be imported
├── src <- Source code for use in this project.
│ ├── __init__.py <- Makes src a Python module
│ │
│ ├── models <- Scripts to create models
Be sure you're connected to the feed and use this command to install TSODA through pip:
make requirements
- Python 3.6, 3.7
- Tensorflow 1.15.2
To check all packages requirements, please see requirements.txt.
- v0.1.0
- bboxes finder
- OCR
You can import a class of bboxes finder and a class of OCR.
⚠️ The following quotations aren't running: You can find in the models folder a lot of scripts about how to train and predict with NNs and extract/preprocess data.
This quotations are here to get a quick view at how to import models and to know how structured input/ output data have to be.
from LRP.models.model_bboxes import PlaqueFinder
from LRP.models.model_OCR import PlaqueOCR#-----------------------------
# Test train PlaqueFinder
#-----------------------------
index = ['train_input', 'the_labels', 'input_length', 'label_length']
X_train = np.ones((1, 128, 64, 3))
Y_train = [[0.9, 0.3, 0.1, 0.1]] # x, y, h, w in % of image height and width
PF = PlaqueFinder(shape=(224, 224, 3), weight=None, loss='iou')
epochs = 200
batch_size = 8
PF.train(x_train=X_train,
y_train=Y_train,
batch_size=batch_size,
epochs=epochs,
validation_data=(X_val, y_val))
#-----------------------------
# Test train OCR
#-----------------------------
from LRP.models.utils import gen_flow_for_two_inputs
index = ['train_input', 'the_labels', 'input_length', 'label_length']
input_train = {'train_input': np.ones((1, 128, 64, 3)),
'the_labels' : [[0, 1, 2]], # For ABC
'input_length': [30],
'label_length': [3]}
output_train = {'ctc': [0]} # Vector of zeros
print('Train a GRU')
POCR = PlaqueOCR(shape=(128, 64, 3),
shapes=[10],
gru=512,
optimizers=Adadelta(lr=0.8))
gen_flow = gen_flow_for_two_inputs(input_train, output_train['ctc'])
POCR.train_generator(gen_flow=gen_flow,
epochs=100,
steps_per_epoch=len(input_train['train_input']) / 32,
validation_data=(input_val, output_val))| Image with context | Resized Image |
|---|---|
 |
 |
Embeding PlaqueFinder with Darflow
| Image with context | ROI Image selected | Resized Image |
|---|---|---|
 |
 |
 |
| Resized Image | Text |
|---|---|
|
PJB7392 |
- Add probability of containing a licence plate
- Create an Autoencoder for resized image denoising (& Centering ?) before OCR
- Docstrings
- Sphinx autodoc
- Unit test