print(time.time() - a)
y_train, y_test = np.take(y,train_index,axis=0), np.take(y,test_index,axis=0)
print("Size training set", len(X_train))
print("Size test set", len(X_test))
if fold_num==1:
file.write("Size training set: %d\n" % len(X_train))
file.write("Size test set: %d\n" % len(X_test))
file.write("Class distribution:\n")
file.write("Train;Test\n")
dtrain = Counter(y_train)
dtest = Counter(y_test)
for i in range(IndianPines_Input.NUM_CLASSES):
file.write(str(i)+";"+ str(dtrain[i]) + ";" + str(dtest[i]) + "\n")
save_path,final_test_acc,conf_matrix = CNNTrain_2D.train_model(X_train, y_train, X_test, y_test, config)
print(final_test_acc)
accuracies.append(final_test_acc)
file.write("Fold "+ str(fold_num) + ";" + "%.3f" % final_test_acc + "\n" )
# Clear memory
del X_train, X_test, y_train, y_test
fold_num+=1
print(accuracies)
mean_acc = np.asarray(accuracies,dtype=float).mean()
file.write("Mean accuracy;" + "%.3f" % mean_acc + "\n\n" )
print("Size test set", len(X_test))
file.write("\n------------------\nResults for patch size " + str(patch_size) + ":\n")
if validation_set:
file.write("Class distribution:\n")
file.write("#;Train;Validation;Test\n")
dtrain = Counter(y_train)
dval = Counter(y_val)
dtest = Counter(y_test)
for i in range(input.num_classes):
file.write(str(i+1)+";" + str(dtrain[i]) + ";" + str(dval[i]) + ";" + str(dtest[i]) + "\n")
#
save_path, val_acc, conf_matrix = CNNTrain_2D.train_model(X_train, y_train, X_val, y_val, config)
# save_path, final_test_acc, conf_matrix = CNNTrain_2D.train_model(X_test, y_test, X_train, y_train, config)
# Clear memory
del X_train, X_val, X_test, y_train, y_val, y_test
else:
file.write("Class distribution:\n")
file.write("#;Train;Test\n")
dtrain = Counter(y_train)
dtest = Counter(y_test)
for i in range(input.num_classes):
file.write(str(i + 1) + ";" + str(dtrain[i]) + ";" + str(dtest[i]) + "\n")
save_path, test_acc, conf_matrix = CNNTrain_2D.train_model(X_train, y_train, X_test, y_test, config)
file.write("Size test set: %d\n" % len(X_test))
file.write("Class distribution:\n")
file.write("Train;Test\n")
dtrain = Counter(y_train)
dtest = Counter(y_test)
for i in range(input.num_classes):
file.write(
str(i) + ";" + str(dtrain[i]) + ";" +
str(dtest[i]) + "\n")
file.write(
"Fold;Train acc; Test acc; Test Post acc;Kappa\n")
print('Start training')
t = time.time()
save_path, test_acc, _ = CNNTrain_2D.train_model(
X_train, y_train, X_test, y_test, config)
t = time.time() - t
print(t)
if fold_num == 1:
file.write("Time %0.3f\n" % t)
print("Test accuracy:" + str(test_acc))
raw, train_acc, test_acc = CV_Decoder.decode_cnn(
input, config, train_positions, test_positions, save_path)
print("Train accuracy:" + str(train_acc))
print("Test accuracy:" + str(test_acc))
save_rgb(fold_dir + "outmap.png",