Spaces:
Runtime error
Runtime error
| import os | |
| import gdown | |
| import yaml | |
| import numpy as np | |
| import uuid | |
| import requests | |
| import tempfile | |
| from tqdm import tqdm | |
| def download_weights(uri, cached=None, md5=None, quiet=False): | |
| if uri.startswith('http'): | |
| return download(url=uri, quiet=quiet) | |
| return uri | |
| def download(url, quiet=False): | |
| tmp_dir = tempfile.gettempdir() | |
| filename = url.split('/')[-1] | |
| full_path = os.path.join(tmp_dir, filename) | |
| if os.path.exists(full_path): | |
| print('Model weight {} exsits. Ignore download!'.format(full_path)) | |
| return full_path | |
| with requests.get(url, stream=True) as r: | |
| r.raise_for_status() | |
| with open(full_path, 'wb') as f: | |
| for chunk in tqdm(r.iter_content(chunk_size=8192)): | |
| # If you have chunk encoded response uncomment if | |
| # and set chunk_size parameter to None. | |
| #if chunk: | |
| f.write(chunk) | |
| return full_path | |
| def download_config(id): | |
| url = 'https://vocr.vn/data/vietocr/config/{}'.format(id) | |
| r = requests.get(url) | |
| config = yaml.safe_load(r.text) | |
| return config | |
| def compute_accuracy(ground_truth, predictions, mode='full_sequence'): | |
| """ | |
| Computes accuracy | |
| :param ground_truth: | |
| :param predictions: | |
| :param display: Whether to print values to stdout | |
| :param mode: if 'per_char' is selected then | |
| single_label_accuracy = correct_predicted_char_nums_of_single_sample / single_label_char_nums | |
| avg_label_accuracy = sum(single_label_accuracy) / label_nums | |
| if 'full_sequence' is selected then | |
| single_label_accuracy = 1 if the prediction result is exactly the same as label else 0 | |
| avg_label_accuracy = sum(single_label_accuracy) / label_nums | |
| :return: avg_label_accuracy | |
| """ | |
| if mode == 'per_char': | |
| accuracy = [] | |
| for index, label in enumerate(ground_truth): | |
| prediction = predictions[index] | |
| total_count = len(label) | |
| correct_count = 0 | |
| try: | |
| for i, tmp in enumerate(label): | |
| if tmp == prediction[i]: | |
| correct_count += 1 | |
| except IndexError: | |
| continue | |
| finally: | |
| try: | |
| accuracy.append(correct_count / total_count) | |
| except ZeroDivisionError: | |
| if len(prediction) == 0: | |
| accuracy.append(1) | |
| else: | |
| accuracy.append(0) | |
| avg_accuracy = np.mean(np.array(accuracy).astype(np.float32), axis=0) | |
| elif mode == 'full_sequence': | |
| try: | |
| correct_count = 0 | |
| for index, label in enumerate(ground_truth): | |
| prediction = predictions[index] | |
| if prediction == label: | |
| correct_count += 1 | |
| avg_accuracy = correct_count / len(ground_truth) | |
| except ZeroDivisionError: | |
| if not predictions: | |
| avg_accuracy = 1 | |
| else: | |
| avg_accuracy = 0 | |
| else: | |
| raise NotImplementedError('Other accuracy compute mode has not been implemented') | |
| return avg_accuracy | |