3588板子部署yoloV5_rk3588非量化部署yolo

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 context
  rga_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;
  
  
}