def test_evaluate_model_with_sampling(self):
c = Params()
c.load_config_file(JSON_PATH)
model_name = "simple_cnn"
sample_name = "4000_256"
imgCone = ImageCone(c, Open_Slide())
# 读取数字全扫描切片图像
tag = imgCone.open_slide("Testing/images/test_001.tif", None,
"test_001")
images = imgCone.get_uniform_sampling_blocks(40, 256)
print("sample count = ", len(images))
normal = ACDNormalization("acd",
dc_txt="dc.txt",
w_txt="w.txt",
template_path="template_normal")
normal.prepare(images)
source_samples = ("P0404", "T_NC_Simple0404_4000_256_test.txt")
cnn = Simple_Classifier(c,
model_name,
sample_name,
normalization=normal,
special_norm=0)
cnn.evaluate_model(samples_name=source_samples,
model_file=None,
batch_size=20,
max_count=None)
def test_evaluate_model3(self):
c = Params()
c.load_config_file(JSON_PATH)
# model_name = "se_densenet_40"
# model_name = "simple_cnn"
# model_name = "resnet_18"
model_name = "densenet_22"
sample_name = "4000_256"
cnn = Simple_Classifier(c,
model_name,
sample_name,
normalization=None,
special_norm=-1)
result = cnn.evaluate_models(
samples_name=("P0404", "T_NC_Simple0404_4000_256_test.txt"),
model_directory="{}/models/pytorch/{}_{}".format(
c.PROJECT_ROOT, model_name, sample_name),
batch_size=60,
max_count=None)
pd.set_option('display.max_colwidth', 300)
pd.set_option('display.max_rows', None)
pd.set_option('display.max_columns', None)
pd.set_option('display.width', 2000)
print(result)
def test_evaluate_model2(self):
c = Params()
c.load_config_file(JSON_PATH)
# model_name = "se_densenet_40"
# sample_name = "x_256"
# model_name = "simple_cnn"
model_name = "e_densenet_22"
sample_name = "4000_256"
normal = None
cnn = Simple_Classifier(c,
model_name,
sample_name,
normalization=normal,
special_norm=-1)
# Check_P0430_4000_256_test,
# T1_P0430_4000_256_train,
# T1_P0430_4000_256_test.txt
Xtest, Ytest, predicted_tags, features = cnn.evaluate_model(
samples_name=("P0430", "T1_P0430_4000_256_train.txt"),
model_file=None,
batch_size=100,
max_count=None)
cnn.evaluate_accuracy_based_slice(Xtest, predicted_tags, Ytest)
def test_export_ONNX_model(self):
c = Params()
c.load_config_file(JSON_PATH)
# model_name = "se_densenet_22"
model_name = "e_densenet_22"
sample_name = "4000_256"
cnn = Simple_Classifier(c, model_name, sample_name)
cnn.export_ONNX_model()
def test_train_model_cifar(self):
c = Params()
c.load_config_file(JSON_PATH)
# model_name = "simple_cnn"
model_name = "e_densenet_40"
sample_name = "cifar10"
cnn = Simple_Classifier(c, model_name, sample_name)
cnn.train_model(samples_name=None,
augment_func=None,
batch_size=32,
epochs=200)
def test_train_model_patholImg(self):
c = Params()
c.load_config_file(JSON_PATH)
mode = 41
if mode == 1:
model_name = "simple_cnn"
sample_name = "4000_256"
elif mode == 2:
model_name = "densenet_22"
sample_name = "4000_256"
elif mode == 3:
model_name = "e_densenet_22"
sample_name = "4000_256"
elif mode == 4:
model_name = "e_densenet_40"
sample_name = "4000_256"
elif mode == 41:
model_name = "e_densenet_40"
sample_name = "2000_256"
elif mode == 5:
model_name = "se_densenet_40"
sample_name = "4000_256"
elif mode == 6:
model_name = "resnet_18"
sample_name = "4000_256"
elif mode == 7:
pass
cnn = Simple_Classifier(c, model_name, sample_name)
samples = [
("P0619", "T1_P0619_4000_256"), # 平衡样本集
("P0619", "S1_P0619_2000_256"), # 平衡样本集
("P0619", "A1_P0619_4k2k_256"), # 平衡样本集
]
# cnn.train_model(samples_name=("P0430","T1_P0430_4000_256"), class_weight=[0.2391, 1.0],
# augment_func = None, batch_size=30, epochs = 10)
# cnn.train_model(samples_name=samples[1], cl ass_weight=None,
# augment_func = None, batch_size=30, epochs = 10)
cnn.train_model_A2(samples_name=samples[1],
augment_func=None,
class_weight=None,
batch_size=30,
loss_weight=0.001,
epochs=10)
def test_predict_on_batch(self):
c = Params()
c.load_config_file(JSON_PATH)
model_name = "simple_cnn"
sample_name = "4000_256"
cnn = Simple_Classifier(c, model_name, sample_name)
imgCone = ImageCone(c, Open_Slide())
# 读取数字全扫描切片图像
tag = imgCone.open_slide("Train_Tumor/Tumor_004.tif", None,
"Tumor_004")
seeds = [(69400, 98400), (70800, 98400), (27200, 113600)] # C, C, S,
result = cnn.predict_on_batch(imgCone, 40, 256, seeds, 2)
print(result)
def test_evaluate_model(self):
c = Params()
c.load_config_file(JSON_PATH)
# model_name = "se_densenet_40"
# sample_name = "x_256"
# model_name = "simple_cnn"
model_name = "e_densenet_22"
sample_name = "4000_256"
# normal = HistNormalization("match_hist", hist_target ="hist_templates_2048.npy",
# hist_source = "hist_source_P0404.npy")
# normal = HistNormalization("match_hist", hist_target = "hist_soures_P0327.npy",
# hist_source = "hist_soures_P0330.npy")
# normal = HSDNormalization("hsd_norm", target_mean=( -0.2574, 0.2353, 0.3893),
# target_std=(0.1860, 0.1884, 0.2482),
# source_mean=(-0.0676, 0.4088, 0.3710),
# source_std=(0.1254, 0.1247, 0.1988))
# normal = ReinhardNormalization("reinhard", target_mean=(72.66, 16.89, -9.979),
# target_std=(13.42, 5.767, 4.891),
# source_mean=(72.45, 17.63, -17.77),
# source_std=(10.77, 7.064, 6.50))
normal = None
# normal = ImageAugmentation(l_range = (0.95, 1.05), a_range = (0.95, 1.05),
# b_range = (0.95, 1.05), constant_range = (-10, 10))
# normal = ACDNormalization_tf("acd", dc_txt="dc.txt", w_txt="w.txt", template_path="template_normal")
# normal = ACDNormalization("acd", dc_txt="dc.txt", w_txt="w.txt", template_path="Tumor_025")
# source_samples = ("P0404", "T_NC_Simple0404_4000_256_test.txt")
# patch_root = c.PATCHS_ROOT_PATH[source_samples[0]]
# sample_filename = source_samples[1]
# train_list = "{}/{}".format(patch_root, sample_filename)
#
# Xtrain, Ytrain = util.read_csv_file(patch_root, train_list)
#
# # prepare
# images = []
# for patch_file in Xtrain:
# img = io.imread(patch_file, as_gray=False)
# images.append(img)
#
# normal.prepare(images)
cnn = Simple_Classifier(c,
model_name,
sample_name,
normalization=normal,
special_norm=-1)
# cnn.evaluate_model(samples_name=("P0330", "T_NC_Simple0330_4000_256_test.txt"),
# model_file=None, batch_size=20, max_count=None)
# cnn.evaluate_model(samples_name=("P0404", "T_NC_Y0404_4000_256_test.txt"),
# model_file=None, batch_size=20, max_count=None)
# cnn.evaluate_model(samples_name=("P0404", "T_NC_W0404_4000_256_test.txt"),
# model_file=None, batch_size=20, max_count=None)
# cnn.evaluate_model(samples_name=("P0330", "T_NC_Simple0330_4000_256_test.txt"),
# model_file=None, batch_size=60, max_count=None)
cnn.evaluate_model(
samples_name=("P0404", "T_NC_Simple0404_4000_256_test.txt"
), # T_NC_Simple0404_4000_256_test
model_file=None,
batch_size=20,
max_count=None)