Tensorflow based utilities used for training convolutional neural networks. The scripts were used in kaggle competition and helped to reach the accurasy of 91% resulting in the 2nd place.
Packages and other software used in the scripts
- tensorflow(>= 2.0.0)
- opencv-python(>= 4.4.0.42)
python3 train.py -h
usage: t.py [-h] [-mp MODEL_PATH] [-dp DATA_PATH] [-ps PIC_SIZE]
[-bs BATCH_SIZE] [-cp CH_PATH] [-sf SAVE_FREQ]
[-sm] [-lw] [-v VERBOSE] [-e EPOCHS]
optional arguments:
-h, --help show this help message and exit
-mp MODEL_PATH, --model_path MODEL_PATH Path to an h5 model file
-dp DATA_PATH, --data_path DATA_PATH Path to a folder with image files
-ps PIC_SIZE, --pic_size PIC_SIZE Square side size to rescale the pictures. Default (300)
-bs BATCH_SIZE, --batch_size BATCH_SIZE Size of the batch to load pictures. Default (30)
-cp CH_PATH, --checkpoint_path CH_PATH Path to save/load checkpoints
-sf SAVE_FREQ, --save_frequency SAVE_FREQ Checkpoint save frequency in batches default (0)
-sm, --save_model Save model file after the training
-lw, --load_weights Whether to load weights from a checkpoint
-v VERBOSE, --verbose VERBOSE Keras verbosity level (0-2)
-e EPOCHS, --epochs EPOCHS Number of epochs to train