3588板子部署yoloV5

一 ：准备 ubuntu linux X86_64系统

        a.安装anaconda

        b.创建虚拟环境      python=3.8

二： 下载rknn-toolkit2

传送门

unzip 解压文件夹   

三：pt转onnx模型

四：onnx转rknn模型

        a:cd到rknn-toolkit2-master/rknn-toolkit2/packages目录下

        cp表示python的版本号  如果不知道版本的话  输入python __version

        b:     pip install  -r requirements_cp38-1.6.0.txt 安装环境

        c:      pip install rknn_toolkit2-1.4.0_22dcfef4-cp38-cp38-linux_x86_64.whl

        d:      from rknn.api import RKNN  如果不报错则安装成功

        e：    去到knn-toolkit2-master/rknn-toolkit2/examples/onnx/yolov5目录下修改test.py

                  修改成自己的

         f:  修改后 运行pyhton test.py  根目录会生成 .rknn的模型以及result.jpg

        

       

 五：3588部署

获取瑞星微官网 rknpu 

         传送门

在model目录下放入推理视频h264或h265格式

./rknn_yolov5_video_demo ./model/RK3588/yolov5n_result.rknn ./model/a.mp4 264或265

至此推理结束

==================================24.3.7更新============================

修改了rknn_yolov5_demo读取视频操作

// Copyright (c) 2021 by Rockchip Electronics Co., Ltd. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License./*-------------------------------------------Includes
-------------------------------------------*/
#include <dlfcn.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>#define _BASETSD_H#include "RgaUtils.h"#include "postprocess.h"#include "rknn_api.h"
#include "preprocess.h"#define PERF_WITH_POST 1
/*-------------------------------------------Functions
-------------------------------------------*/static void dump_tensor_attr(rknn_tensor_attr *attr)
{std::string shape_str = attr->n_dims < 1 ? "" : std::to_string(attr->dims[0]);for (int i = 1; i < attr->n_dims; ++i){shape_str += ", " + std::to_string(attr->dims[i]);}printf("  index=%d, name=%s, n_dims=%d, dims=[%s], n_elems=%d, size=%d, w_stride = %d, size_with_stride=%d, fmt=%s, ""type=%s, qnt_type=%s, ""zp=%d, scale=%f\n",attr->index, attr->name, attr->n_dims, shape_str.c_str(), attr->n_elems, attr->size, attr->w_stride,attr->size_with_stride, get_format_string(attr->fmt), get_type_string(attr->type),get_qnt_type_string(attr->qnt_type), attr->zp, attr->scale);
}double __get_us(struct timeval t) { return (t.tv_sec * 1000000 + t.tv_usec); }static unsigned char *load_data(FILE *fp, size_t ofst, size_t sz)
{unsigned char *data;int ret;data = NULL;if (NULL == fp){return NULL;}ret = fseek(fp, ofst, SEEK_SET);if (ret != 0){printf("blob seek failure.\n");return NULL;}data = (unsigned char *)malloc(sz);if (data == NULL){printf("buffer malloc failure.\n");return NULL;}ret = fread(data, 1, sz, fp);return data;
}static unsigned char *load_model(const char *filename, int *model_size)
{FILE *fp;unsigned char *data;fp = fopen(filename, "rb");if (NULL == fp){printf("Open file %s failed.\n", filename);return NULL;}fseek(fp, 0, SEEK_END);int size = ftell(fp);data = load_data(fp, 0, size);fclose(fp);*model_size = size;return data;
}static int saveFloat(const char *file_name, float *output, int element_size)
{FILE *fp;fp = fopen(file_name, "w");for (int i = 0; i < element_size; i++){fprintf(fp, "%.6f\n", output[i]);}fclose(fp);return 0;
}/*-------------------------------------------Main Functions
-------------------------------------------*/
int main(int argc, char **argv)
{if (argc < 2){printf("Usage: %s <rknn model> <input_image_path> <resize/letterbox> <output_image_path>\n", argv[0]);return -1;}int ret;rknn_context ctx;size_t actual_size = 0;int img_width = 0;int img_height = 0;int img_channel = 0;const float nms_threshold = NMS_THRESH;      // 默认的NMS阈值const float box_conf_threshold = BOX_THRESH; // 默认的置信度阈值struct timeval start_time, stop_time;char *model_name = (char *)argv[1];//char *input_path = argv[2];std::string option = "letterbox";std::string out_path = "./out.jpg";if (argc >= 4){option = argv[3];}if (argc >= 5){out_path = argv[4];}// init rga contextrga_buffer_t src;rga_buffer_t dst;memset(&src, 0, sizeof(src));memset(&dst, 0, sizeof(dst));printf("post process config: box_conf_threshold = %.2f, nms_threshold = %.2f\n", box_conf_threshold, nms_threshold);/* Create the neural network */printf("Loading mode...\n");int model_data_size = 0;unsigned char *model_data = load_model(model_name, &model_data_size);ret = rknn_init(&ctx, model_data, model_data_size, 0, NULL);if (ret < 0){printf("rknn_init error ret=%d\n", ret);return -1;}rknn_sdk_version version;ret = rknn_query(ctx, RKNN_QUERY_SDK_VERSION, &version, sizeof(rknn_sdk_version));if (ret < 0){printf("rknn_init error ret=%d\n", ret);return -1;}printf("sdk version: %s driver version: %s\n", version.api_version, version.drv_version);rknn_input_output_num io_num;ret = rknn_query(ctx, RKNN_QUERY_IN_OUT_NUM, &io_num, sizeof(io_num));if (ret < 0){printf("rknn_init error ret=%d\n", ret);return -1;}printf("model input num: %d, output num: %d\n", io_num.n_input, io_num.n_output);rknn_tensor_attr input_attrs[io_num.n_input];memset(input_attrs, 0, sizeof(input_attrs));for (int i = 0; i < io_num.n_input; i++){input_attrs[i].index = i;ret = rknn_query(ctx, RKNN_QUERY_INPUT_ATTR, &(input_attrs[i]), sizeof(rknn_tensor_attr));if (ret < 0){printf("rknn_init error ret=%d\n", ret);return -1;}dump_tensor_attr(&(input_attrs[i]));}rknn_tensor_attr output_attrs[io_num.n_output];memset(output_attrs, 0, sizeof(output_attrs));for (int i = 0; i < io_num.n_output; i++){output_attrs[i].index = i;ret = rknn_query(ctx, RKNN_QUERY_OUTPUT_ATTR, &(output_attrs[i]), sizeof(rknn_tensor_attr));dump_tensor_attr(&(output_attrs[i]));}int channel = 3;int width = 0;int height = 0;if (input_attrs[0].fmt == RKNN_TENSOR_NCHW){printf("model is NCHW input fmt\n");channel = input_attrs[0].dims[1];height = input_attrs[0].dims[2];width = input_attrs[0].dims[3];}else{printf("model is NHWC input fmt\n");height = input_attrs[0].dims[1];width = input_attrs[0].dims[2];channel = input_attrs[0].dims[3];}printf("model input height=%d, width=%d, channel=%d\n", height, width, channel);rknn_input inputs[1];memset(inputs, 0, sizeof(inputs));inputs[0].index = 0;inputs[0].type = RKNN_TENSOR_UINT8;inputs[0].size = width * height * channel;inputs[0].fmt = RKNN_TENSOR_NHWC;inputs[0].pass_through = 0;// 读取图片//printf("Read %s ...\n", input_path);cv::VideoCapture cap("/home/pi/rknpu2-master/examples/rknn_yolov5_demo/model/sample.mp4");cv::Mat frame;printf("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa");while(cap.read(frame)){//cv::imshow("frame",frame);cv::Mat orig_img = frame.clone();//cv::Mat orig_img = cv::imread(input_path, 1);if (!orig_img.data){printf("cv::imread fail!\n");return -1;}cv::Mat img;cv::cvtColor(orig_img, img, cv::COLOR_BGR2RGB);img_width = img.cols;img_height = img.rows;printf("img width = %d, img height = %d\n", img_width, img_height);// 指定目标大小和预处理方式,默认使用LetterBox的预处理BOX_RECT pads;memset(&pads, 0, sizeof(BOX_RECT));cv::Size target_size(width, height);cv::Mat resized_img(target_size.height, target_size.width, CV_8UC3);// 计算缩放比例float scale_w = (float)target_size.width / img.cols;float scale_h = (float)target_size.height / img.rows;if (img_width != width || img_height != height){// 直接缩放采用RGA加速if (option == "resize"){printf("resize image by rga\n");ret = resize_rga(src, dst, img, resized_img, target_size);if (ret != 0){fprintf(stderr, "resize with rga error\n");return -1;}// 保存预处理图片cv::imwrite("resize_input.jpg", resized_img);}else if (option == "letterbox"){printf("resize image with letterbox\n");float min_scale = std::min(scale_w, scale_h);scale_w = min_scale;scale_h = min_scale;letterbox(img, resized_img, pads, min_scale, target_size);// 保存预处理图片cv::imwrite("letterbox_input.jpg", resized_img);}else{fprintf(stderr, "Invalid resize option. Use 'resize' or 'letterbox'.\n");return -1;}inputs[0].buf = resized_img.data;}else{inputs[0].buf = img.data;}gettimeofday(&start_time, NULL);rknn_inputs_set(ctx, io_num.n_input, inputs);rknn_output outputs[io_num.n_output];memset(outputs, 0, sizeof(outputs));for (int i = 0; i < io_num.n_output; i++){outputs[i].want_float = 0;}// 执行推理ret = rknn_run(ctx, NULL);ret = rknn_outputs_get(ctx, io_num.n_output, outputs, NULL);gettimeofday(&stop_time, NULL);printf("once run use %f ms\n", (__get_us(stop_time) - __get_us(start_time)) / 1000);// 后处理detect_result_group_t detect_result_group;std::vector<float> out_scales;std::vector<int32_t> out_zps;for (int i = 0; i < io_num.n_output; ++i){out_scales.push_back(output_attrs[i].scale);out_zps.push_back(output_attrs[i].zp);}post_process((int8_t *)outputs[0].buf, (int8_t *)outputs[1].buf, (int8_t *)outputs[2].buf, height, width,box_conf_threshold, nms_threshold, pads, scale_w, scale_h, out_zps, out_scales, &detect_result_group);// 画框和概率char text[256];for (int i = 0; i < detect_result_group.count; i++){detect_result_t *det_result = &(detect_result_group.results[i]);sprintf(text, "%s %.1f%%", det_result->name, det_result->prop * 100);printf("%s @ (%d %d %d %d) %f\n", det_result->name, det_result->box.left, det_result->box.top,det_result->box.right, det_result->box.bottom, det_result->prop);int x1 = det_result->box.left;int y1 = det_result->box.top;int x2 = det_result->box.right;int y2 = det_result->box.bottom;rectangle(orig_img, cv::Point(x1, y1), cv::Point(x2, y2), cv::Scalar(256, 0, 0, 256), 3);putText(orig_img, text, cv::Point(x1, y1 + 12), cv::FONT_HERSHEY_SIMPLEX, 0.4, cv::Scalar(255, 255, 255));}printf("save detect result to %s\n", out_path.c_str());imwrite(out_path, orig_img);imshow("aaa",orig_img);cv::waitKey(1);ret = rknn_outputs_release(ctx, io_num.n_output, outputs);// 耗时统计//int test_count = 10;//gettimeofday(&start_time, NULL);//for (int i = 0; i < test_count; ++i)//{//rknn_inputs_set(ctx, io_num.n_input, inputs);// ret = rknn_run(ctx, NULL);// ret = rknn_outputs_get(ctx, io_num.n_output, outputs, NULL);
//#if PERF_WITH_POST//   post_process((int8_t *)outputs[0].buf, (int8_t *)outputs[1].buf, (int8_t *)outputs[2].buf, height, width,//                box_conf_threshold, nms_threshold, pads, scale_w, scale_h, out_zps, out_scales, &detect_result_group);
//#endif// ret = rknn_outputs_release(ctx, io_num.n_output, outputs);//}//gettimeofday(&stop_time, NULL);//printf("loop count = %d , average run  %f ms\n", test_count,//     (__get_us(stop_time) - __get_us(start_time)) / 1000.0 / test_count);//deinitPostProcess();// release//ret = rknn_destroy(ctx);// if (model_data)// {//  free(model_data);//}}return 0;}