]
slide_list_3 = [
95, 120, 81, 77, 97, 96, 110, 83, 152, 128, 149, 155, 153, 111, 57, 138,
134, 135, 114, 76, 123, 90, 121, 61, 147, 148, 119, 142, 66, 137, 63, 80,
70, 79, 115, 133, 129, 141
]
# ALL SAMPLE DATA
data_start_time = datetime.now()
all_samples = pd.DataFrame(columns=columns)
# negative samples add
for i in range(len(slide_list_1)):
image_path = image_paths[slide_list_1[i]]
mask_path = tumor_mask_paths[slide_list_1[i]]
samples = find_patches_from_slide(image_path, mask_path)
samples = samples.sample(2000, random_state=42, replace=True)
all_samples = all_samples.append(samples)
for i in range(len(slide_list_2)):
image_path = image_paths[slide_list_2[i]]
mask_path = tumor_mask_paths[slide_list_2[i]]
samples = find_patches_from_slide(image_path, mask_path)
tumor_samples = samples[samples.is_tumor == True]
tumor_samples = tumor_samples.sample(1000, random_state=42, replace=True)
non_tumor_samples = samples[samples.is_tumor == False]
non_tumor_samples = non_tumor_samples.sample(1000,
random_state=42,
replace=True)
samples = tumor_samples.append(non_tumor_samples)
all_samples = all_samples.append(samples)
for i in range(len(slide_list_3)):
PATCH_SIZE = 256
BATCH_SIZE = 128
test_image_paths = read_test_data_path_2()
print(test_image_paths)
start_x = 64
start_y = 64
pred_size = 128
slide_id = list()
slide_pred = list()
for id_test in range(len(test_image_paths)):
print(id_test, 'th inference\n')
image_path = test_image_paths[id_test]
test_samples = find_patches_from_slide(image_path, 'test')
NUM_SAMPLES = len(test_samples)
if NUM_SAMPLES > 5000:
NUM_SAMPLES = 5000
samples = test_samples.sample(NUM_SAMPLES, random_state=42)
samples.reset_index(drop=True, inplace=True)
test_generator = gen_imgs_test(image_path, 'test', samples, BATCH_SIZE)
test_steps = np.ceil(len(samples) / BATCH_SIZE)
preds = []
for i in range(int(test_steps)):
X, Y = next(test_generator)
for j in range(len(X)):
