Raviv Shalev | 97ddc06 | 2021-12-07 15:18:09 +0200 | [diff] [blame] | 1 | # Copyright © 2022 Arm Ltd and Contributors. All rights reserved. |
| 2 | # SPDX-License-Identifier: MIT |
| 3 | |
| 4 | import os |
| 5 | import cv2 |
| 6 | import numpy as np |
| 7 | |
| 8 | from context import style_transfer |
| 9 | from context import cv_utils |
| 10 | |
| 11 | |
| 12 | def test_style_transfer_postprocess(test_data_folder): |
| 13 | content_image = "messi5.jpg" |
| 14 | target_shape = (1,256,256,3) |
| 15 | keep_aspect_ratio = False |
| 16 | image = cv2.imread(os.path.join(test_data_folder, content_image)) |
| 17 | original_shape = image.shape |
| 18 | preprocessed_image = cv_utils.preprocess(image, np.float32, target_shape, False, keep_aspect_ratio) |
| 19 | assert preprocessed_image.shape == target_shape |
| 20 | |
| 21 | postprocess_image = style_transfer.style_transfer_postprocess(preprocessed_image, original_shape) |
| 22 | assert postprocess_image.shape == original_shape |
| 23 | |
| 24 | |
| 25 | def test_style_transfer(test_data_folder): |
| 26 | style_predict_model_path = os.path.join(test_data_folder, "style_predict.tflite") |
| 27 | style_transfer_model_path = os.path.join(test_data_folder, "style_transfer.tflite") |
| 28 | backends = ["CpuAcc", "CpuRef"] |
| 29 | delegate_path = os.path.join(test_data_folder, "libarmnnDelegate.so") |
| 30 | image = cv2.imread(os.path.join(test_data_folder, "messi5.jpg")) |
| 31 | |
| 32 | style_transfer_executor = style_transfer.StyleTransfer(style_predict_model_path, style_transfer_model_path, |
| 33 | image, backends, delegate_path) |
| 34 | |
| 35 | assert style_transfer_executor.get_style_predict_executor_shape() == (1, 256, 256, 3) |
| 36 | |
| 37 | def test_run_style_transfer(test_data_folder): |
| 38 | style_predict_model_path = os.path.join(test_data_folder, "style_predict.tflite") |
| 39 | style_transfer_model_path = os.path.join(test_data_folder, "style_transfer.tflite") |
| 40 | backends = ["CpuAcc", "CpuRef"] |
| 41 | delegate_path = os.path.join(test_data_folder, "libarmnnDelegate.so") |
| 42 | style_image = cv2.imread(os.path.join(test_data_folder, "messi5.jpg")) |
| 43 | content_image = cv2.imread(os.path.join(test_data_folder, "basketball1.png")) |
| 44 | |
| 45 | style_transfer_executor = style_transfer.StyleTransfer(style_predict_model_path, style_transfer_model_path, |
| 46 | style_image, backends, delegate_path) |
| 47 | |
| 48 | stylized_image = style_transfer_executor.run_style_transfer(content_image) |
| 49 | assert stylized_image.shape == content_image.shape |
| 50 | |
| 51 | |
| 52 | def test_create_stylized_detection(test_data_folder): |
| 53 | style_predict_model_path = os.path.join(test_data_folder, "style_predict.tflite") |
| 54 | style_transfer_model_path = os.path.join(test_data_folder, "style_transfer.tflite") |
| 55 | backends = ["CpuAcc", "CpuRef"] |
| 56 | delegate_path = os.path.join(test_data_folder, "libarmnnDelegate.so") |
| 57 | |
| 58 | style_image = cv2.imread(os.path.join(test_data_folder, "messi5.jpg")) |
| 59 | content_image = cv2.imread(os.path.join(test_data_folder, "basketball1.png")) |
| 60 | detections = [(0.0, [0.16745174, 0.15101701, 0.5371381, 0.74165875], 0.87597656)] |
| 61 | labels = {0: ('person', (50.888902345757494, 129.61878417939724, 207.2891028294508)), |
| 62 | 1: ('bicycle', (55.055339686943654, 55.828708219750574, 43.550389695374676)), |
| 63 | 2: ('car', (95.33096265662336, 194.872841553212, 218.58516479057758))} |
| 64 | style_transfer_executor = style_transfer.StyleTransfer(style_predict_model_path, style_transfer_model_path, |
| 65 | style_image, backends, delegate_path) |
| 66 | |
| 67 | stylized_image = style_transfer.create_stylized_detection(style_transfer_executor, 'person', content_image, |
| 68 | detections, 720, labels) |
| 69 | |
| 70 | assert stylized_image.shape == content_image.shape |