| import cv2 |
| import numpy as np |
| |
|
|
|
|
| def read_image(file_bytes): |
| np_img = np.frombuffer(file_bytes, np.uint8) |
| img = cv2.imdecode(np_img, cv2.IMREAD_COLOR) |
| return img |
|
|
| def combine_mask(image, mask, label2color={1: (255, 255, 0), 2: (0, 255, 255)}, alpha=0.1): |
| image = to_color(image) |
| image = cv2.resize(image, mask.shape) |
| mask_image = np.zeros_like(image) |
| for label, color in label2color.items(): |
| mask_image[mask == label] = color |
|
|
| mask_image = cv2.addWeighted(image, 1 - alpha, mask_image, alpha, 0) |
| return mask_image |
|
|
|
|
| def to_color(image): |
| if len(image.shape) == 3 and image.shape[-1] == 3: |
| return image |
| return cv2.cvtColor(image, cv2.COLOR_GRAY2RGB) |
|
|
|
|
| def to_gray(image): |
| if len(image.shape) == 3 and image.shape[-1] == 3: |
| return cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) |
| return image |
|
|
| def prob_process(prob_list): |
| prob = prob_list[0] |
| max_idx = np.argmax(prob) |
| min_idx = np.argmin(prob) |
| prob[max_idx] += prob[min_idx] |
| prob[min_idx] = 0 |
| |
| if max_idx == 0: |
| return 1 - prob[max_idx] |
| elif max_idx ==1: |
| return prob[max_idx] |
| else: |
| return 1 + prob[max_idx] |
|
|
| def combine_prob_jsw(prob, jsw_m, jsw_mm): |
| processed_prob = prob_process(prob) |
| temp = np.array([processed_prob, jsw_m, jsw_mm]) |
| if temp.any(): |
| temp = np.hstack(temp) |
| else: |
| temp = temp.flatten() |
| |
| return temp |
|
|
| def scale_coordinates(links, original_size, mask_size): |
| scale_factor = original_size / mask_size |
| scaled_links = [(int(x * scale_factor), int(y * scale_factor)) for x, y in links] |
| return scaled_links |
|
|
| def get_annotations(probabilites): |
| OSTEOPYTE_LEVELS = { |
| 0: "Definite osteophytes", |
| 1: "No osteophytes", |
| 2: "Possible osteophytes", |
| } |
|
|
| JNS_LEVELS = { |
| 0: "Definite JSN", |
| 1: "Mild JSN", |
| 2: "No JSN", |
| 3: "Severe JSN", |
| } |
|
|
| return { |
| "osteophyte": OSTEOPYTE_LEVELS[np.argmax(probabilites[5:8])], |
| "jsn": JNS_LEVELS[np.argmax(probabilites[8:12])], |
| } |
|
|
