def inference(params):
config_gpu() # 配置GPU环境
# 构建模型
print("创建模型 ...")
model = Seq2Seq(params=params)
# 获取保存管理者
print("创建模型保存器")
checkpoint = tf.train.Checkpoint(Seq2Seq=model)
checkpoint_manager = tf.train.CheckpointManager(checkpoint,
config.checkpoint_path,
max_to_keep=3)
if checkpoint_manager.latest_checkpoint:
print("加载最新保存器数据 {} ...".format(checkpoint_manager.latest_checkpoint))
checkpoint.restore(checkpoint_manager.latest_checkpoint)
# 训练模型
print("开始预测 ...")
inference_model(model, params)
def train(params):
config_gpu() # 配置GPU环境
# 构建模型
print("创建模型 ...")
model = seq2seq(params=params)
# 获取保存管理者
print("创建模型保存器")
checkpoint = tf.train.Checkpoint(Seq2Seq=model)
checkpoint_manager = tf.train.CheckpointManager(checkpoint,
config.checkpoint_path,
max_to_keep=3)
if checkpoint_manager.latest_checkpoint:
print("加载最新保存器数据 {} ...".format(checkpoint_manager.latest_checkpoint))
checkpoint.restore(checkpoint_manager.latest_checkpoint)
else:
print("初始化保存器.")
# 训练模型
print("开始训练 ...")
train_model(model, params, checkpoint_manager)
def beam_search(params):
config_gpu() # 配置GPU
# 加载数据集、模型
_, _, test_X = load_data()
model = Seq2Seq(params)
start = time.time()
print('使用集束搜索开始预测...')
results = []
dataset, steps_per_epoch = test_batch_generator(params['batch_size'])
with tqdm(total=steps_per_epoch, position=0, leave=True) as tq:
for (batch, batch_x) in enumerate(dataset.take(steps_per_epoch)):
results += beam_decode(model, params, batch_x)
tq.update(1)
print('预测完成，耗时{}s\n处理至文件...'.format(time.time() - start))
def result_proc(text):
"""
对预测结果做最后处理
:param text: 单条预测结果
:return:
"""
# text = text.lstrip(' ，！。')
text = text.replace(' ', '')
text = text.strip()
if '<end>' in text:
text = text[:text.index('<end>')]
return text
test_csv = pd.read_csv(config.test_set, encoding="UTF-8")
# 赋值结果
test_csv['Prediction'] = results
# 提取ID和预测结果两列
test_df = test_csv[['QID', 'Prediction']]
# 结果处理
test_df['Prediction'] = test_df['Prediction'].apply(result_proc)
# 保存结果
test_df.to_csv(config.inference_result_path, index=None, sep=',')
print('已保存文件至{}'.format(config.inference_result_path))
self.dataset["annotations"].append({
"id":
image_id,
"image_id":
image_id,
"category_id":
cid,
"bbox": [x1, y1, x2 - x1, y2 - y1],
"area": (x2 - x1) * (y2 - y1),
"score":
1,
"iscrowd":
False
})
def evaluating(cocoGt, cocoDt):
handler = COCOeval(cocoGt, cocoDt, "bbox")
handler.evaluate()
handler.accumulate()
handler.summarize()
if __name__ == "__main__":
config_gpu()
params = build_params()
result = run_batch(params)
gt = GestureEval(None, params=params)
dt = GestureEval(result, params)
evaluating(gt, dt)