test_time_aug

eval.py

@@ -0,0 +1,123 @@
+import os
+os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"
+os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
+os.environ["CUDA_VISIBLE_DEVICES"] = "0"
+
+import tensorflow as tf
+import numpy as np
+import cv2
+import pandas as pd
+from glob import glob
+from tqdm import tqdm
+import tensorflow as tf
+from tensorflow.keras.utils import CustomObjectScope
+from sklearn.metrics import accuracy_score, f1_score, jaccard_score, precision_score, recall_score
+from metrics import dice_loss, dice_coef, iou
+from train import create_dir, load_data
+
+H = 256
+W = 256
+
+def read_image(path):
+    x = cv2.imread(path, cv2.IMREAD_COLOR)  ## (H, W, 3)
+    x = cv2.resize(x, (W, H))
+    ori_x = x
+    x = x/255.0
+    x = x.astype(np.float32)
+    x = np.expand_dims(x, axis=0)
+    return ori_x, x                                ## (1, 256, 256, 3)
+
+
+def read_mask(path):
+    x = cv2.imread(path, cv2.IMREAD_GRAYSCALE)  ## (H, W)
+    x = cv2.resize(x, (W, H))
+    ori_x = x
+    x = x/255.0
+    x = x.astype(np.int32)                    ## (256, 256)
+    return ori_x, x
+
+def save_results(ori_x, ori_y, y_pred, save_image_path):
+    line = np.ones((H, 10, 3)) * 255
+
+    ori_y = np.expand_dims(ori_y, axis=-1)  ## (256, 256, 1)
+    ori_y = np.concatenate([ori_y, ori_y, ori_y], axis=-1) ## (256, 256, 3)
+
+    y_pred = np.expand_dims(y_pred, axis=-1)  ## (256, 256, 1)
+    y_pred = np.concatenate([y_pred, y_pred, y_pred], axis=-1) ## (256, 256, 3)
+
+    cat_images = np.concatenate([ori_x, line, ori_y, line, y_pred*255], axis=1)
+    cv2.imwrite(save_image_path, cat_images)
+
+if __name__ == "__main__":
+    """ Seeding """
+    np.random.seed(42)
+    tf.random.set_seed(42)
+
+    """ Folder for saving results """
+    create_dir("results_DLV3SA_2value_no_early_stop")
+
+    """ Load the model """
+    with CustomObjectScope({'iou': iou, 'dice_coef': dice_coef, 'dice_loss': dice_loss}):
+        model = tf.keras.models.load_model("files_DLV3SA/model_DLV3SA_2value_no_early_stop.h5")
+
+    """ Load the test data """
+    dataset_path = "D:/medical_challenge/segmentation/CVC-612/data/"
+    # dataset_path = "D:/medical_challenge/segmentation/2d/VIP_CUP_5fold/fold_1/train/"
+    # dataset_path = "D:/medical_challenge/segmentation/CVC-612/CVC_ClinicDB"
+    # dataset_path = "D:/medical_challenge/segmentation/2d/data_mbrain_5fold/fold_0/train/"
+    (train_x, train_y), (valid_x, valid_y) = load_data(dataset_path)
+    # print(len(test_x), len(test_y))
+    print("=====================================================================================================")
+    print(len(valid_x), len(valid_y))
+    print("=====================================================================================================")
+    print(len(train_x), len(train_y))
+    SCORE = []
+    for x, y in tqdm(zip(valid_x, valid_y), total=len(valid_x)):
+        """ Exctracting the image name """
+        # name = x.split("/")[-1]
+        # print(x)
+        name = x
+        """ Read the image and mask """
+        ori_x, x = read_image(x)
+        ori_y, y = read_mask(y)
+
+        """ Predicting the mask """
+        y_pred = model.predict(x)[0] > 0.5
+        y_pred = np.squeeze(y_pred, axis=-1)
+        y_pred = y_pred.astype(np.int32)
+
+        """ Saving the predicted mask """
+        save_image_path = os.path.join(os.getcwd(), 'results_DLV3SA_2value_no_early_stop', os.path.basename(name))
+        ## os.getcwd: lấy path fiel hiện tại
+        ## basename: lấy tên ảnh
+
+        # print("*"*50, save_image_path)
+        save_results(ori_x, ori_y, y_pred, save_image_path)
+
+        """ Flatten the array """
+        y = y.flatten()
+        y_pred = y_pred.flatten()
+
+        """ Calculating metrics values """
+        # dice_coef_value = 1 - dice_loss(y, y_pred)
+        acc_value = accuracy_score(y, y_pred)
+        f1_value = f1_score(y, y_pred, labels=[0, 1], average="binary")
+        jac_value = jaccard_score(y, y_pred, labels=[0, 1], average="binary")
+        recall_value = recall_score(y, y_pred, labels=[0, 1], average="binary")
+        precision_value = precision_score(y, y_pred, labels=[0, 1], average="binary")
+        SCORE.append([name, acc_value, f1_value, jac_value, recall_value, precision_value])
+
+
+    """ mean metrics values """
+    score = [s[1:] for s in SCORE]
+    score = np.mean(score, axis=0)
+    # print(f"Dice score: {score[0]:0.5f}")
+    print(f"Accuracy: {score[0]:0.5f}")
+    print(f"F1: {score[1]:0.5f}")
+    print(f"Jaccard: {score[2]:0.5f}")
+    print(f"Recall: {score[3]:0.5f}")
+    print(f"Precision: {score[4]:0.5f}")
+
+    df = pd.DataFrame(SCORE, columns = ["Image Name", "Acc", "F1", "Jaccard", "Recall", "Precision"])
+    df.to_csv("files_DLV3SA/score_DLV3SA_2value_no_early_stop.csv")
+