Model is trained using SSD_MobileNet_V1 architecture cascading to OCR for automatically recognise number plates of Indian vehicles.

Dataset used: https://www.kaggle.com/dataturks/vehicle-number-plate-detection Developed using Tensorflow 1.15

• Data-Images.zip: Contains the images of the cars, number plates and annotations in .txt files (YOLO format)
• Data_preparation.ipynb: A notebook demonstrating the process of preparing the dataset (.csv files) for creating TFRecords (otherwise TensorFlow Object Detection API won't work)
• Detection.ipynb: A notebook demonstrating the process of detecting number plates from video feed and crop the image to be sent for OCR (this notebook has to be stored inside object_detection folder)
• Recognition.ipynb: A notebook demonstrating the process of recognising digits of the number plate, generated from Detection.ipynb
• Indian_Number_plates.json: Configuration file which contains image download paths and annotations
• exported_graph: Contains the inference graph in .pb and .tflite formats which can be used to run inference on both CPU platforms and on-device platforms
• label_map.pbtxt: Contains the encodings of the dataset classes which,in this case, is 1: license_plate
• ssd_mobilenet_v1_pets.config: Training and evaluation pipeline configuration file as needed by TensorFlow Object Detection API
• test.record & train.record: TFRecords files of testing and training sets respectively
• test_labels.csv & train_labels.csv: .csv files as required by the generate_tfrecord.py script
• requirements.txt: Python dependencies to install

To kick-start the model training process, Run the train file located in object_detection/legacy/ python train.py –-logtostderr –train_sir=training/ --pipeline_config_path=training/ssd_mobilenet_v1_pets.config

To save frozen graph for inference, Run the following command (To be able convert an inference graph to its .tflite variant you need to enable quantization aware training and you can specify that in the .config file itself.) :

tflite_convert \
    --output_file=detect.tflite \
    --graph_def_file=frozen_inference_graph.pb \
    --input_shapes=1,300,300,3 \
    --input_arrays=normalized_input_image_tensor \
    --output_arrays='TFLite_Detection_PostProcess','TFLite_Detection_PostProcess:1','TFLite_Detection_PostProcess:2','TFLite_Detection_PostProcess:3'  \
    --inference_type=QUANTIZED_UINT8 \
    --mean_values=128 \
    --std_dev_values=128 \
    --change_concat_input_ranges=false \
    --allow_custom_ops

Note: A collection of pre-trained detection models are avaiable here, if you want to train it with another model.