# 랜덤하게 이미지를 한장 뽑아내는 함수
np.random.seed = 100
def random_test_image():
"""Pick a random test image from the test directory"""
c = np.random.choice(cat_df['category'])
root = testdir + c + '/'
img_path = root + np.random.choice(os.listdir(root))
return img_path
# 랜덤하게 이미지를 선택한 후 Top5 예측치를 확인하는 함수
train_util.display_prediction(random_test_image(),
model,
topk=5,
model_name=model_choice)
time.sleep(1.5)
train_util.display_prediction(random_test_image(),
model,
topk=5,
model_name=model_choice)
time.sleep(1.5)
train_util.display_prediction(random_test_image(),
model,
topk=5,
model_name=model_choice)
time.sleep(1.5)
train_util.display_prediction(random_test_image(),
model,
topk=5,
# 랜덤하게 이미지를 한장 뽑아내는 함수
np.random.seed = 100
def random_test_image():
"""Pick a random test image from the test directory"""
c = np.random.choice(cat_df['category'])
root = testdir + c + '/'
img_path = root + np.random.choice(os.listdir(root))
return img_path
avg_inference_time = 0
# 랜덤하게 이미지를 선택한 후 Top5 예측치를 확인하는 함수
avg_inference_time += train_util.display_prediction(random_test_image(),
model,
topk=5,
model_name=model_choice)
time.sleep(1.5)
avg_inference_time += train_util.display_prediction(random_test_image(),
model,
topk=5,
model_name=model_choice)
time.sleep(1.5)
avg_inference_time += train_util.display_prediction(random_test_image(),
model,
topk=5,
model_name=model_choice)
time.sleep(1.5)
avg_inference_time += train_util.display_prediction(random_test_image(),
model,
topk=5,
# 랜덤하게 이미지를 한장 뽑아내는 함수
np.random.seed = 100
def random_test_image():
"""Pick a random test image from the test directory"""
c = np.random.choice(cat_df['category'])
root = testdir + c + '/'
img_path = root + np.random.choice(os.listdir(root))
return img_path
avg_inference_time = 0
# 랜덤하게 이미지를 선택한 후 Top5 예측치를 확인하는 함수
for a in range(10):
avg_inference_time += train_util.display_prediction(
random_test_image(),
model,
topk=5,
model_name=model_choice,
etc='[K_' + str(K_fold) + '_' + str(k) + '_fold] ')
time.sleep(1.1)
total_avg_inference_time += avg_inference_time
print(f'총 10회 추론시간 평균 : {avg_inference_time*1000.0/10:.2f}ms')
results.append(
train_util.evaluate(model, dataloaders['test'], criterion, n_classes))
results[k] = results[k].merge(
cat_df, left_on='class',
right_on='category').drop(columns=['category'])
results[k].sort_values('top1', ascending=False, inplace=True)