paddleocr-tensorrt加速-Serving部署

1
2
 load_static_weights: True                 #是否将预训练模型保存在静态图形模式

1
2
3
4
5
6
pip install paddle-serving-server==0.6.1 # for CPU
pip install paddle-serving-server-gpu==0.6.1 # for GPU
# 其他GPU环境需要确认环境再选择执行如下命令
pip install paddle-serving-server-gpu==0.6.1.post101 # GPU with CUDA10.1 + TensorRT6
pip install paddle-serving-server-gpu==0.6.1.post11 # GPU with CUDA11 + TensorRT7

1
2
3
wget https://paddle-serving.bj.bcebos.com/test-dev/whl/paddle_serving_client-0.6.1-cp37-none-any.whl
pip install paddle_serving_client-0.6.1-cp37-none-any.whl

1
2
3
wget https://paddle-serving.bj.bcebos.com/test-dev/whl/paddle_serving_app-0.6.1-py3-none-any.whl
pip install paddle_serving_app-0.6.1-py3-none-any.whl

1
2
3
4
5
6
7
8
9
10
11
12
13
14
# 转换检测模型
python3 -m paddle_serving_client.convert --dirname ./ch_PP-OCRv2_det_infer/  # 使用自己训练好的模型路径
                                         --model_filename inference.pdmodel          
                                         --params_filename inference.pdiparams       
                                         --serving_server ./ppocrv2_det_serving/   # 保存位置可以更改为指定位置
                                         --serving_client ./ppocrv2_det_client/  # 保存位置可以更改为指定位置

# 转换识别模型  同上
python3 -m paddle_serving_client.convert --dirname ./ch_PP-OCRv2_rec_infer/ 
                                         --model_filename inference.pdmodel          
                                         --params_filename inference.pdiparams       
                                         --serving_server ./ppocrv2_rec_serving/  
                                         --serving_client ./ppocrv2_rec_client/

1
2
3
4
class RecOp(Op):
    def init_op(self):
    	self.det_box = []

1
2
3
4
5
6
7
8
9
10
def preprocess(self, input_dicts, data_id, log_id):
	 self.det_box = []
	        for i in dt_boxes:
	            b = []
	            for j in i:
	                x, y = j
	                b.append([int(x), int(y)])
	            # print("bb:", b)
	            self.det_box.append(b)

1
2
3
def postprocess(self, input_dicts, fetch_data, data_id, log_id):
	box = str(self.det_box)