run_cli.py

import argparse
from src.learner import BoWLearner
from src.utils import load_json
import cv2
import os


def train(config_path):
    print(f"Load config from {config_path}.")
    config = load_json(config_path)
    learner = BoWLearner.from_config(config=config)
    learner.train()


def evaluate(serialization_dir, test_path=None, result_path='results'):
    learner = BoWLearner.from_serialization_dir(serialization_dir=serialization_dir)
    learner.evaluate(img_paths=test_path, result_path=result_path)


def infer(serialization_dir, image_path, imshow=False, img_size=(480, 580)):
    learner = BoWLearner.from_serialization_dir(serialization_dir=serialization_dir)
    y_prediction = learner.predict(image_path, imshow=imshow, img_size=img_size)
    print('Prediction: ', y_prediction)


if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--mode', type=str, default=None)
    parser.add_argument('--config_path',  type=str, default='configs/natural_image_config.json')
    parser.add_argument('--serialization_dir', type=str, default=None)
    parser.add_argument('--result_path', type=str, default='results')
    parser.add_argument('--test_path', type=str, default='data/natural_images/test_names.csv')
    parser.add_argument('--image_path', type=str, default='data/natural_images/car/car_0000.jpg')
    parser.add_argument('--imshow', type=bool, default=False)

    args = parser.parse_args()
    if args.mode == 'train':
        train(config_path=args.config_path)
    elif args.mode == 'eval':
        if args.serialization_dir is None:
            config = load_json(args.config_path)
            print(f"Load config from {args.config_path}.")
            serialization_dir = config['serialization_dir']
            config = load_json(args.config_path)
        else:
            serialization_dir = args.serialization_dir

        if args.test_path is None:
            config = load_json(args.config_path)
            print(f"Load test_path from config {args.config_path}.")
            test_path = config['test_path']
        else:
            test_path = args.test_path

        evaluate(serialization_dir, test_path, args.result_path)

    elif args.mode == 'infer':
        if args.serialization_dir is None:
            config = load_json(args.config_path)
            print(f"Load config from {args.config_path}.")
            serialization_dir = config['serialization_dir']
        else:
            serialization_dir = args.serialization_dir
        infer(serialization_dir, image_path=args.image_path, imshow=args.imshow)