C++ 部署接口的执行效率
已关闭
C++ 部署接口的执行效率
Created by: pele228
windows下编译的C++预测接口测试,1080卡+CUDA10, yolov3_r34模型处理单帧平均时长为 608608 80ms 416416 65ms 320320 62ms yolov3_mobilenet_v1模型 608608 65ms 416416 55ms 320320 52ms
yolov3_mobilenet_v1_voc.yaml文件如下 DEPLOY: USE_GPU: 1 MODEL_PATH: "yolov3_mobilenet_v1_voc" MODEL_FILENAME: "model" PARAMS_FILENAME: "params" EVAL_CROP_SIZE: (320, 320) RESIZE_TYPE: "UNPADDING" MEAN: [0.485, 0.456, 0.406] STD: [0.229, 0.224, 0.225] IMAGE_TYPE: "rgb" NUM_CLASSES: 26 CHANNELS : 3 PRE_PROCESSOR: "DetectionPreProcessor" PREDICTOR_MODE: "ANALYSIS" BATCH_SIZE : 1 RESIZE_MAX_SIZE: 1333 FEEDS_SIZE: 3
感觉执行效率很低, 还需要对参数设置做哪些调整吗?
Created by: qingqing01
已经重构了C++预测代码,支持使用预测库的优化,以及TensorRT https://github.com/PaddlePaddle/PaddleDetection/tree/release/0.3/deploy