feature_extractor = Model(inputs=feature_extractor.input, outputs=x)
rm = ResultManager('evaluation_results')
agent_acc_size_dict = []
origin_acc_size_dict = []
agent = DQN_Agent(s_dim=1280,
a_dim=10,
epsilon_decay=0.99,
epsilon_min=0.02,
gamma=0.95,
replay_batchsize=256)
if EVALUATION:
agent.model = load_model(
'evaluation_results/agent_DQN_train_amazon_imagenet.h5'
) # If in evaluation phase, replace this with the actual pretrained model
agent.curr_exploration_rate = 0
step_count = 0
env = EnvironmentAPI(
imagenet_train_path=images_dir,
# cloud_agent=AmazonRekognition(),
cloud_agent=FacePP(),
dataset='imagenet',
# cache_path='evaluation_results/image_reference_cache_amazon.defaultdict')
cache_path='evaluation_results/image_reference_cache_face.defaultdict')
# In order to reduce some billing recognition service requests, I cached the recognized result locally.
# Can be replaced by loading an empty dict from a pickled file. Navigate to the code for more details.
test_image_paths = imagenet_paths[:10]
print(test_image_paths)
flag = 1
for index, path in enumerate(test_image_paths):
image = Image.open(path).convert("RGB")
image_data = preprocess_input(
np.expand_dims(np.asarray(image.resize((224, 224)),
dtype=np.float32),
axis=0))
# features = feature_extractor.predict(image_data)[0][0][0]
# state_actions, action_id = agent.choose_action(features)
# action = [i for i in np.arange(5, 105, 10)][action_id]
print(index)
print(array)
# time.sleep(1)
if flag == 1:
print("xxx")
print(array)
handle_process.join()
agent.model = array[0]
flag = 0
# if __name__ == "__main__":
# main()
# array = multiprocessing.Manager().list()
# print('父进程pid: %d' % os.getpid()) # os.getpid获取当前进程的进程号
# handle_process = Process(target=process_handle, args=(array,))
# handle_process.start()
# # handle_process.join()
# array.extend([5, 6, 7])
# print(array)
# test_image_paths = FLIR[:1000]
cloud_agent = Baidu()
train_log = defaultdict(list)
ref_results = defaultdict(dict)
compress_results = defaultdict(dict)
choose_action_total_time = 0
feedback_total_time = 0
feature_total_time = 0
recent_accuracy = 0
recent_reward = 0
start_time = time.time()
if EVALUATION:
agent.model = load_model('compute_time_results/baidu_imagenet.h5')
# agent.model = load_model('compute_time_results/baidu_FLIR.h5')
agent.curr_exploration_rate = 0
load_model_time = time.time()
# init feature
compute_memory(103)
image = Image.open(test_image_paths[0]).convert("RGB")
image_data = preprocess_input(
np.expand_dims(np.asarray(image.resize((224, 224)), dtype=np.float32),
axis=0))
features = feature_extractor.predict(image_data)[0][0][0]
compute_memory(104)
# for
for step_count, path in enumerate(test_image_paths):
compute_memory(step_count)
image = Image.open(path).convert("RGB")