上传文件至 src

README.md

@@ -1,6 +1,8 @@
-# rknn_yolo_EAI_pic
+# 本串口协议必备要求：先发送在接收
 
-### 作者：崔志佳
+## 注：本项目已经通过串口测试（传感器为115200，摄像头为9500），每次报警前会识别设备名称序号，然后反馈到串口输出序列中
+
+# rknn_yolo_EAI_pic
 
 #### 简要说明
 
@@ -8,7 +10,7 @@
 集成了从串口读取红外温度数据，到达阈值后报警并运行rknn模型
 
 需注意运行该例程需要将npu驱动更新为1.7.3版本
- 
+
 本项目只适用于EAI-YOLOV5
 详情请参考EAI官网：https://www.easy-eai.com/document_details/3/342
 

include/jlinux_uart.h

@@ -0,0 +1,81 @@
+/***********************************************************************
+ * @file jlinux_uart.h
+     JLINUX_UART
+ * @brief   header file
+ * @history
+ * Date       Version Author    description
+ * ========== ======= ========= =======================================
+ * 2022-07-27 V1.0    Lucky,lukai@jovision.com   Create
+ *
+ * @Copyright (C)  2022  Jovision Technology Co., Ltd.
+***********************************************************************/
+#ifndef __JLINUX_UART_H__
+#define __JLINUX_UART_H__
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif
+
+typedef struct _uart_ctx *juart_hdl_t;
+
+typedef struct
+{
+	int baudrate;   //波特率:1200/2400/4800/9600/19200/38400/57600/115200/230400/380400/460800/921600
+	int datawidth;  //数据位宽度:5/6/7/8
+	int stopbit;    //停止位宽度:1/2
+	int parity;		//奇偶校验:0无校验，1奇校验，2偶校验
+}JUartAttr_t;
+
+/**
+ *@brief 打开串口设备设备，如果是485设备请配合jctrl_rs485相关接口使用
+ *@param name 设备名称，如：/dev/ttyS0
+ *@return 句柄
+ */
+juart_hdl_t juart_open(const char *name);
+
+/**
+ *@brief 关闭串口
+ *@param handle 句柄
+ */
+int juart_close(juart_hdl_t handle);
+
+/**
+ *@brief 配置串口属性
+ *@param handle 句柄
+ *@param attr 属性
+ */
+int juart_set_attr(juart_hdl_t handle, JUartAttr_t *attr);
+
+int juart_get_fd(juart_hdl_t uart);
+/**
+ *@brief 通过串口发送数据
+ *@param handle 句柄
+ *@param data 数据buffer
+ *@param len 数据长度
+ *@return 0成功
+ */
+int juart_send(juart_hdl_t handle, char *data, int len);
+
+/**
+ *@brief 通过串口接受数据
+ *@param handle 句柄
+ *@param data 数据buffer
+ *@param len buffer的长度
+ *@param timeout 超时毫秒
+ *@return 收到数据的长度
+ */
+int juart_recv(juart_hdl_t handle, char *data, int len, int timeout);
+
+/**
+ *@brief 设置rs485模式
+ *@param handle 句柄
+ *@param mode 0:发送后立刻置为接收状态；非0：发送后保持发送状态
+ *@return 0 成功
+ */
+int juart_set_rs485(juart_hdl_t handle, int mode);
+
+#ifdef __cplusplus
+}
+#endif
+#endif // __JLINUX_UART_H__

src/jlinux_uart.cpp

@@ -0,0 +1,313 @@
+/***********************************************************************
+ * @file jctrl_uart.cpp
+     JCTRL_UART
+ * @brief   header file
+ * @history
+ * Date       Version Author    description
+ * ========== ======= ========= =======================================
+ * 2022-07-21 V1.0    Lucky,lukai@jovision.com   Create
+ *
+ * @Copyright (C)  2022  Jovision Technology Co., Ltd.
+***********************************************************************/
+#include <termios.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <unistd.h>
+#include <errno.h>
+#include <linux/stat.h>
+#include <sys/prctl.h>
+#include <sys/ioctl.h>
+#include <sys/types.h>
+#include <fcntl.h>
+#include <linux/serial.h>
+#include "jlinux_uart.h"
+
+struct _uart_ctx{
+    int fd;
+};
+
+int _get_baudrate(int nBaud)
+{
+    switch(nBaud)
+    {
+    case 1200:
+        return B1200;	//注：B1200为系统定义
+    case 2400:
+        return B2400;
+    case 4800:
+        return B4800;
+    case 9600:
+        return B9600;
+ 	case 19200 :
+        return B19200;
+	case 38400:
+	     return B38400;
+	case 57600:
+	     return B57600;
+	case 115200:
+	     return B115200;
+    default:
+        return B2400;
+    }
+}
+
+
+juart_hdl_t juart_open(const char *name){
+    juart_hdl_t uart = new _uart_ctx;
+    uart->fd = open(name, O_RDWR | O_NONBLOCK | O_NOCTTY |  O_EXCL|O_SYNC);
+    return uart;
+}
+int juart_close(juart_hdl_t uart){
+    if(uart->fd>0)
+		close(uart->fd);
+	uart->fd = 0;
+    return 0;
+}
+int juart_get_fd(juart_hdl_t uart){
+    return uart->fd;
+}
+int juart_set_attr(juart_hdl_t uart, JUartAttr_t *attr){
+    if (uart->fd <= 0)
+    {
+        printf("jv_uart_recv_ex fd error\n");
+        return -1;
+    }
+    struct termios newtio, oldtio;
+    memset(&oldtio, 0, sizeof(oldtio));
+    /* save the old serial port configuration */
+    if (tcgetattr(uart->fd, &oldtio) != 0) {
+        perror("set_port/tcgetattr");
+        return -1;
+    }
+    memset(&newtio, 0, sizeof(newtio));
+    //设置波特率
+    int nBaud = _get_baudrate(attr->baudrate);
+    switch (nBaud)
+    {
+    case B300:
+    case B1200:
+    case B2400:
+    case B4800:
+    case B9600:
+    case B19200:
+    case B38400:
+    case B57600:
+    case B115200:
+        cfsetospeed(&newtio, nBaud);
+        cfsetispeed(&newtio, nBaud);
+        break;
+    default:
+        printf("jv_uart_set_attr:Unsupported baudrate!\n");
+        return -1;
+    }
+    /* ignore modem control lines and enable receiver */
+    newtio.c_cflag |= CLOCAL | CREAD;
+    newtio.c_cflag &= ~CSIZE;
+    /* set character size */
+    switch (attr->datawidth) {
+    case 5:
+        newtio.c_cflag |= CS5;
+        break;
+    case 6:
+        newtio.c_cflag |= CS6;
+        break;
+    case 7:
+        newtio.c_cflag |= CS7;
+        break;
+    case 8:
+    default:
+        newtio.c_cflag |= CS8;
+        break;
+    }
+    /* set the stop bits */
+    switch (attr->stopbit) {
+    default:
+    case 1:
+        newtio.c_cflag &= ~CSTOPB;
+        break;
+    case 2:
+        newtio.c_cflag |= CSTOPB;
+        break;
+    }
+    /* set the parity */
+    switch (attr->parity) {
+    case 'o':
+    case 'O':
+    case 1:
+        newtio.c_cflag |= PARENB;
+        newtio.c_cflag |= PARODD;
+        newtio.c_iflag |= INPCK;
+        break;
+    case 'e':
+    case 'E':
+    case 2:
+        newtio.c_cflag |= PARENB;
+        newtio.c_cflag &= ~PARODD;
+        newtio.c_iflag |= INPCK;
+        break;
+    case 'n':
+    case 'N':
+    case 0:
+    default:
+        newtio.c_cflag &= ~PARENB;
+        newtio.c_iflag &= ~INPCK;
+        break;
+    }
+    /* Raw input */
+    newtio.c_lflag &= ~(ICANON | ECHO | ECHOE | ISIG);
+    /* Software flow control is disabled */
+    newtio.c_iflag &= ~(IXON | IXOFF | IXANY);
+    /* Raw ouput */
+    newtio.c_oflag &=~ OPOST;
+    /* set timeout in deciseconds for non-canonical read */
+    newtio.c_cc[VTIME] = 0;
+    /* set minimum number of characters for non-canonical read */
+    newtio.c_cc[VMIN] = 0;
+    /* flushes data received but not read */
+    tcflush(uart->fd, TCIFLUSH);
+    /* set the parameters associated with the terminal from
+        the termios structure and the change occurs immediately */
+    if ((tcsetattr(uart->fd, TCSANOW, &newtio)) != 0) {
+        perror("set_port/tcsetattr");
+        return -1;
+    }
+    return 0;
+}
+int juart_send(juart_hdl_t uart, char *data, int len){
+    if (uart->fd > 0)
+    {
+        int ret = write(uart->fd, data, len);
+        if (ret == len)
+            return 0;
+    }
+    return -1;
+}
+int _modbus_rtu_select(juart_hdl_t uart, struct timeval *tv)
+{
+    fd_set rfds;
+    FD_ZERO(&rfds);
+    FD_SET(uart->fd, &rfds);
+    int s_rc;
+    while ((s_rc = select(uart->fd+1, &rfds, NULL, NULL, tv)) == -1) {
+        if (errno == EINTR) {
+            fprintf(stderr, "A non blocked signal was caught\n");
+            /* Necessary after an error */
+            FD_ZERO(&rfds);
+            FD_SET(uart->fd, &rfds);
+        } else {
+            return -1;
+        }
+    }
+    if (s_rc == 0) {
+        /* Timeout */
+        errno = ETIMEDOUT;
+        return -1;
+    }
+    return s_rc;
+}
+int juart_recv(juart_hdl_t uart, char *data, int len, int timeout){
+    if (uart->fd > 0)
+    {
+        struct timeval tv;
+        tv.tv_sec = 0;
+        tv.tv_usec = timeout*1000;
+        if(_modbus_rtu_select(uart, &tv) > 0){
+            return read(uart->fd, data, len);
+        }
+        return -1;
+    }else{
+        usleep(timeout*1000);
+    }
+    return -1;
+
+}
+
+
+/**
+ *@brief 接收串口数据扩展接口，从收到start开始接收，到收到stop返回
+ *@param handle 句柄
+ *@param data 接收数据缓存
+ *@param len 缓存大小
+ *@param start 数据头标志：前nstart个字节
+ *@param nstart 数据头字节数
+ *@param stop 数据截止标志：后nstop个字节
+ *@param nstop 数据截止标志的字节数
+ */
+extern "C" int juart_recv_ex(juart_hdl_t handle, char *data, int len, char *start, int nstart, char *stop, int nstop, int timeout);
+int juart_recv_ex(juart_hdl_t uart, char *data, int len, char *start, int nstart, char *stop, int nstop, int timeout){
+    if (uart->fd <= 0)
+    {
+        printf("jv_uart_recv_ex fd error\n");
+        return -1;
+    }
+    int offset = 0;
+    int offset_end = 0;
+    int bytes_read = 0;
+    while (1)
+    {
+        bytes_read = read(uart->fd, &data[offset], 1);
+        if (bytes_read == 1 && offset < nstart && data[offset] == start[offset])
+        {
+            offset++;
+        }
+        if (offset == nstart)
+            break;
+        if (bytes_read < 1)
+            usleep(0);
+    }
+    while (offset < len)
+    {
+        if (data[offset] == stop[offset_end])
+        {
+            offset_end++;
+        }
+        if (offset_end == nstop)
+            break;
+        bytes_read = read(uart->fd, &data[offset], 1);
+        if (bytes_read == 1)
+        {
+            offset++;
+        }
+        else
+        {
+            usleep(0);
+        }
+    }
+    return offset;
+}
+
+/**
+ *@brief 设置rs485模式
+ *@param handle 句柄
+ *@param mode 0:发送后立刻置为接收状态；非0：发送后保持发送状态
+ *@return 0 成功
+ */
+int juart_set_rs485(juart_hdl_t handle, int mode)
+{
+    struct serial_rs485 rs485;
+	if (ioctl(handle->fd, TIOCGRS485, &rs485) == -1)
+	{
+		printf("TIOCGRS485 ioctl error.\n");
+		return -1;
+	}
+	rs485.flags |= SER_RS485_ENABLED;
+	if (mode == 0)
+	{
+		rs485.flags &= ~SER_RS485_RTS_ON_SEND;
+		rs485.flags |= SER_RS485_RTS_AFTER_SEND;
+	}
+	else
+	{
+		rs485.flags |= SER_RS485_RTS_ON_SEND;
+		rs485.flags &= ~SER_RS485_RTS_AFTER_SEND;
+	}
+	rs485.delay_rts_before_send = 0;
+	rs485.delay_rts_after_send = 0;
+	if (ioctl(handle->fd, TIOCSRS485, &rs485) == -1)
+	{
+	    printf("TIOCSRS485 ioctrl error.\n");
+		return -1;
+	}
+	return 0;
+}

src/main.h

@@ -62,6 +62,8 @@ void *upload_message_controller(void *args);
 void *heart_beat(void *args);     //上传心跳检测
 void *distortion(void *args);      //矫正
 void *read_serial_thread(void *args); //读取串口传来的红外温度数据
+void *storage_serial_thread(void *args);
+
 
 struct Alarm {
     int ifalarm;

src/yolov5_detect.cpp

@@ -1,374 +1,374 @@
-#include <iostream>
-#include <fstream>
-#include <vector>
-#include <cstdint>
-#include <stdio.h>
-#include <stdlib.h>
-#include <math.h>
-#include <unistd.h>
-#include <time.h>
-#include <math.h>
-#include <fcntl.h>
-#include <opencv2/opencv.hpp>
-#include "yolov5_detect.h"
-#include "rknn_api.h"
-
-#include <sys/time.h>
-
-using namespace std;
-using namespace cv;
-
-
-//unsigned char *model;
-//detection* dets;
-
-static void printRKNNTensor(rknn_tensor_attr *attr)
-{
-    printf("index=%d name=%s n_dims=%d dims=[%d %d %d %d] n_elems=%d size=%d "
-           "fmt=%d type=%d qnt_type=%d fl=%d zp=%d scale=%f\n",
-           attr->index, attr->name, attr->n_dims, attr->dims[3], attr->dims[2],
-           attr->dims[1], attr->dims[0], attr->n_elems, attr->size, 0, attr->type,
-           attr->qnt_type, attr->fl, attr->zp, attr->scale);
-}
-
-static int letter_box(cv::Mat input_image, cv::Mat *output_image, int model_input_size)
-{
-	int input_width, input_height;
-
-	input_width = input_image.cols;
-	input_height = input_image.rows;
-	float ratio;
-	ratio = min((float)model_input_size / input_width, (float)model_input_size / input_height);
-
-	int new_width, new_height;
-	new_width = round(ratio * input_width );
-	new_height = round(ratio * input_height);
-
-
-	int height_padding = 0;
-	int width_padding = 0;
-	int top = 0;
-	int bottom = 0;
-	int left = 0;
-	int right = 0;
-	if( new_width >= new_height)
-	{
-		height_padding = new_width - new_height;
-		if( (height_padding % 2) == 0 )
-		{
-			top = (int)((float)(height_padding/2));
-			bottom = (int)((float)(height_padding/2));
-		}
-		else
-		{
-			top = (int)((float)(height_padding/2));
-			bottom = (int)((float)(height_padding/2))+1;	
-		}
-	}
-	else
-	{
-		width_padding = new_height - new_width;
-		if( (width_padding % 2) == 0 )
-		{
-			left = (int)((float)(width_padding/2));
-			right = (int)((float)(width_padding/2));
-		}
-		else
-		{
-			left = (int)((float)(width_padding/2));
-			right = (int)((float)(width_padding/2))+1;
-		}
-
-	}
-
-	cv::Mat resize_img;
-
-	cv::resize(input_image, resize_img, cv::Size(new_width, new_height));
-	cv::copyMakeBorder(resize_img, *output_image, top, bottom, left, right, cv::BORDER_CONSTANT, cv::Scalar(0, 0, 0));
-
-	return 0;
-}
-
-int yolov5_detect_init(rknn_context *ctx, const char * path)
-{
-	int ret;
-
-	// Load model
-	FILE *fp = fopen(path, "rb");
-	if(fp == NULL)
-	{
-		printf("fopen %s fail!\n", path);
-		return -1;
-	}
-	fseek(fp, 0, SEEK_END);   //fp指向end,fseek(FILE *stream, long offset, int fromwhere);
-	int model_len = ftell(fp);   //相对文件首偏移
-	unsigned char *model_data = (unsigned char*)malloc(model_len);
-
-	fseek(fp, 0, SEEK_SET);   //SEEK_SET为文件头
-	if(model_len != fread(model_data, 1, model_len, fp))
-	{
-		printf("fread %s fail!\n", path);
-		free(model_data);
-		return -1;
-	}
-	fclose(fp);
-
-	//init
-	ret = rknn_init(ctx, model_data, model_len, RKNN_FLAG_PRIOR_MEDIUM);
-	if(ret < 0)
-	{
-		printf("rknn_init fail! ret=%d\n", ret);
-		return -1;
-	}
-
-	free(model_data);
-
-	return 0;
-}
-
-static int scale_coords(yolov5_detect_result_group_t *detect_result_group, int img_width, int img_height, int model_size)
-{
-	for (int i = 0; i < detect_result_group->count; i++)
-	{
-		yolov5_detect_result_t *det_result = &(detect_result_group->results[i]);
-
-
-		int x1 = det_result->box.left;
-		int y1 = det_result->box.top;
-		int x2 = det_result->box.right;
-		int y2 = det_result->box.bottom;
-
-		
-		if( img_width >= img_height )
-		{
-			int image_max_len = img_width;
-			float gain;
-			gain = (float)model_size / image_max_len;
-			int resized_height = img_height * gain;
-			int height_pading = (model_size - resized_height)/2;
-			y1 = (y1 - height_pading);
-			y2 = (y2 - height_pading);
-			x1 = int(x1 / gain);
-			y1 = int(y1 / gain);
-			x2 = int(x2 / gain);
-			y2 = int(y2 / gain);
-
-			det_result->box.left = x1;
-			det_result->box.top = y1;
-			det_result->box.right = x2;
-			det_result->box.bottom = y2;
-		}
-		else
-		{
-			int image_max_len = img_height;
-			float gain;
-			gain = (float)model_size / image_max_len;
-			int resized_width = img_width * gain;
-			int width_pading = (model_size - resized_width)/2;
-			x1 = (x1 - width_pading);
-			x2 = (x2 - width_pading);
-			x1 = int(x1 / gain);
-			y1 = int(y1 / gain);
-			x2 = int(x2 / gain);
-			y2 = int(y2 / gain);
-
-			det_result->box.left = x1;
-			det_result->box.top = y1;
-			det_result->box.right = x2;
-			det_result->box.bottom = y2;	
-		}
-		
-	}
-
-	return 0;
-}
-
-
-int yolov5_detect_run(rknn_context ctx, cv::Mat input_image, yolov5_detect_result_group_t *detect_result_group)
-{
-	int img_width = 0;
-	int img_height = 0;
-	int img_channel = 0;
-
-	size_t actual_size = 0;
-	const float vis_threshold = 0.1;
-	const float nms_threshold = 0.5;
-	const float conf_threshold = 0.2;
-	int ret;
-
-	img_width = input_image.cols;
-	img_height = input_image.rows;
-
-	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;
-		}
-		//printRKNNTensor(&(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));
-		//printRKNNTensor(&(output_attrs[i]));
-	}
-
-	int input_channel = 3;
-	int input_width = 0;
-	int input_height = 0;
-	if (input_attrs[0].fmt == RKNN_TENSOR_NCHW)
-	{
-		//printf("model is NCHW input fmt\n");
-		input_width = input_attrs[0].dims[0];
-		input_height = input_attrs[0].dims[1];
-	}
-	else
-	{
-		//printf("model is NHWC input fmt\n");
-		input_width = input_attrs[0].dims[1];
-		input_height = input_attrs[0].dims[2];
-	}
-
-	/*
-	printf("model input height=%d, width=%d, channel=%d\n", height, width,
-		   channel);
-	*/
-
-	/* Init input tensor */
-	rknn_input inputs[1];
-	memset(inputs, 0, sizeof(inputs));
-	inputs[0].index = 0;
-	inputs[0].type = RKNN_TENSOR_UINT8;
-	inputs[0].size = input_width * input_height * input_channel;
-	inputs[0].fmt = RKNN_TENSOR_NHWC;
-	inputs[0].pass_through = 0;
-
-	/* Init output tensor */
-	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;
-	}
-
-	cv::Mat letter_image;
-	letter_box(input_image, &letter_image, input_width);
-	inputs[0].buf = letter_image.data;
-
-	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);
-
-	// Post process
-	std::vector<float> out_scales;
-	std::vector<uint8_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);
-	}
-
-
-	yolov5_post_process_u8((uint8_t *)outputs[0].buf, (uint8_t *)outputs[1].buf, (uint8_t *)outputs[2].buf, input_height, input_width,
-					   conf_threshold, nms_threshold, out_zps, out_scales, detect_result_group);
-
-
-	/*
-	yolov5_post_process_fp((float *)outputs[0].buf, (float *)outputs[1].buf, (float *)outputs[2].buf, input_height, input_width,
-			            conf_threshold, nms_threshold, &detect_result_group);
-	*/
-
-	rknn_outputs_release(ctx, io_num.n_output, outputs);
-
-	scale_coords(detect_result_group, img_width, img_height, input_width);
-
-	return 0;
-}
-
-int yolov5_detect_release(rknn_context ctx)
-{
-    rknn_destroy(ctx);
-	return 0;
-}
-
-
-std::string base64_encode(unsigned char const* bytes_to_encode, unsigned int in_len) {
-  std::string ret;
-  int i = 0;
-  int j = 0;
-  unsigned char char_array_3[3];
-  unsigned char char_array_4[4];
-
-  while (in_len--) {
-    char_array_3[i++] = *(bytes_to_encode++);
-    if (i == 3) {
-      char_array_4[0] = (char_array_3[0] & 0xfc) >> 2;
-      char_array_4[1] = ((char_array_3[0] & 0x03) << 4) + ((char_array_3[1] & 0xf0) >> 4);
-      char_array_4[2] = ((char_array_3[1] & 0x0f) << 2) + ((char_array_3[2] & 0xc0) >> 6);
-      char_array_4[3] = char_array_3[2] & 0x3f;
-
-      for(i = 0; (i <4) ; i++)
-        ret += base64_chars[char_array_4[i]];
-      i = 0;
-    }
-  }
-
-  if (i) {
-    for(j = i; j < 3; j++)
-      char_array_3[j] = '\0';
-
-    char_array_4[0] = (char_array_3[0] & 0xfc) >> 2;
-    char_array_4[1] = ((char_array_3[0] & 0x03) << 4) + ((char_array_3[1] & 0xf0) >> 4);
-    char_array_4[2] = ((char_array_3[1] & 0x0f) << 2) + ((char_array_3[2] & 0xc0) >> 6);
-    char_array_4[3] = char_array_3[2] & 0x3f;
-
-    for (j = 0; (j < i + 1); j++)
-      ret += base64_chars[char_array_4[j]];
-
-    while((i++ < 3))
-      ret += '=';
-  }
-
-  return ret;
-}
-
-static inline bool is_base64(unsigned char c) {
-  return (isalnum(c) || (c == '+') || (c == '/'));
+#include <iostream>
+#include <fstream>
+#include <vector>
+#include <cstdint>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <unistd.h>
+#include <time.h>
+#include <math.h>
+#include <fcntl.h>
+#include <opencv2/opencv.hpp>
+#include "yolov5_detect.h"
+#include "rknn_api.h"
+
+#include <sys/time.h>
+
+using namespace std;
+using namespace cv;
+
+
+//unsigned char *model;
+//detection* dets;
+
+static void printRKNNTensor(rknn_tensor_attr *attr)
+{
+    printf("index=%d name=%s n_dims=%d dims=[%d %d %d %d] n_elems=%d size=%d "
+           "fmt=%d type=%d qnt_type=%d fl=%d zp=%d scale=%f\n",
+           attr->index, attr->name, attr->n_dims, attr->dims[3], attr->dims[2],
+           attr->dims[1], attr->dims[0], attr->n_elems, attr->size, 0, attr->type,
+           attr->qnt_type, attr->fl, attr->zp, attr->scale);
+}
+
+static int letter_box(cv::Mat input_image, cv::Mat *output_image, int model_input_size)
+{
+	int input_width, input_height;
+
+	input_width = input_image.cols;
+	input_height = input_image.rows;
+	float ratio;
+	ratio = min((float)model_input_size / input_width, (float)model_input_size / input_height);
+
+	int new_width, new_height;
+	new_width = round(ratio * input_width );
+	new_height = round(ratio * input_height);
+
+
+	int height_padding = 0;
+	int width_padding = 0;
+	int top = 0;
+	int bottom = 0;
+	int left = 0;
+	int right = 0;
+	if( new_width >= new_height)
+	{
+		height_padding = new_width - new_height;
+		if( (height_padding % 2) == 0 )
+		{
+			top = (int)((float)(height_padding/2));
+			bottom = (int)((float)(height_padding/2));
+		}
+		else
+		{
+			top = (int)((float)(height_padding/2));
+			bottom = (int)((float)(height_padding/2))+1;	
+		}
+	}
+	else
+	{
+		width_padding = new_height - new_width;
+		if( (width_padding % 2) == 0 )
+		{
+			left = (int)((float)(width_padding/2));
+			right = (int)((float)(width_padding/2));
+		}
+		else
+		{
+			left = (int)((float)(width_padding/2));
+			right = (int)((float)(width_padding/2))+1;
+		}
+
+	}
+
+	cv::Mat resize_img;
+
+	cv::resize(input_image, resize_img, cv::Size(new_width, new_height));
+	cv::copyMakeBorder(resize_img, *output_image, top, bottom, left, right, cv::BORDER_CONSTANT, cv::Scalar(0, 0, 0));
+
+	return 0;
+}
+
+int yolov5_detect_init(rknn_context *ctx, const char * path)
+{
+	int ret;
+
+	// Load model
+	FILE *fp = fopen(path, "rb");
+	if(fp == NULL)
+	{
+		printf("fopen %s fail!\n", path);
+		return -1;
+	}
+	fseek(fp, 0, SEEK_END);   //fp指向end,fseek(FILE *stream, long offset, int fromwhere);
+	int model_len = ftell(fp);   //相对文件首偏移
+	unsigned char *model_data = (unsigned char*)malloc(model_len);
+
+	fseek(fp, 0, SEEK_SET);   //SEEK_SET为文件头
+	if(model_len != fread(model_data, 1, model_len, fp))
+	{
+		printf("fread %s fail!\n", path);
+		free(model_data);
+		return -1;
+	}
+	fclose(fp);
+
+	//init
+	ret = rknn_init(ctx, model_data, model_len, RKNN_FLAG_PRIOR_MEDIUM);
+	if(ret < 0)
+	{
+		printf("rknn_init fail! ret=%d\n", ret);
+		return -1;
+	}
+
+	free(model_data);
+
+	return 0;
+}
+
+static int scale_coords(yolov5_detect_result_group_t *detect_result_group, int img_width, int img_height, int model_size)
+{
+	for (int i = 0; i < detect_result_group->count; i++)
+	{
+		yolov5_detect_result_t *det_result = &(detect_result_group->results[i]);
+
+
+		int x1 = det_result->box.left;
+		int y1 = det_result->box.top;
+		int x2 = det_result->box.right;
+		int y2 = det_result->box.bottom;
+
+		
+		if( img_width >= img_height )
+		{
+			int image_max_len = img_width;
+			float gain;
+			gain = (float)model_size / image_max_len;
+			int resized_height = img_height * gain;
+			int height_pading = (model_size - resized_height)/2;
+			y1 = (y1 - height_pading);
+			y2 = (y2 - height_pading);
+			x1 = int(x1 / gain);
+			y1 = int(y1 / gain);
+			x2 = int(x2 / gain);
+			y2 = int(y2 / gain);
+
+			det_result->box.left = x1;
+			det_result->box.top = y1;
+			det_result->box.right = x2;
+			det_result->box.bottom = y2;
+		}
+		else
+		{
+			int image_max_len = img_height;
+			float gain;
+			gain = (float)model_size / image_max_len;
+			int resized_width = img_width * gain;
+			int width_pading = (model_size - resized_width)/2;
+			x1 = (x1 - width_pading);
+			x2 = (x2 - width_pading);
+			x1 = int(x1 / gain);
+			y1 = int(y1 / gain);
+			x2 = int(x2 / gain);
+			y2 = int(y2 / gain);
+
+			det_result->box.left = x1;
+			det_result->box.top = y1;
+			det_result->box.right = x2;
+			det_result->box.bottom = y2;	
+		}
+		
+	}
+
+	return 0;
+}
+
+
+int yolov5_detect_run(rknn_context ctx, cv::Mat input_image, yolov5_detect_result_group_t *detect_result_group)
+{
+	int img_width = 0;
+	int img_height = 0;
+	int img_channel = 0;
+
+	size_t actual_size = 0;
+	const float vis_threshold = 0.1;
+	const float nms_threshold = 0.5;
+	const float conf_threshold = 0.2;
+	int ret;
+
+	img_width = input_image.cols;
+	img_height = input_image.rows;
+
+	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;
+		}
+		//printRKNNTensor(&(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));
+		//printRKNNTensor(&(output_attrs[i]));
+	}
+
+	int input_channel = 3;
+	int input_width = 0;
+	int input_height = 0;
+	if (input_attrs[0].fmt == RKNN_TENSOR_NCHW)
+	{
+		//printf("model is NCHW input fmt\n");
+		input_width = input_attrs[0].dims[0];
+		input_height = input_attrs[0].dims[1];
+	}
+	else
+	{
+		//printf("model is NHWC input fmt\n");
+		input_width = input_attrs[0].dims[1];
+		input_height = input_attrs[0].dims[2];
+	}
+
+	/*
+	printf("model input height=%d, width=%d, channel=%d\n", height, width,
+		   channel);
+	*/
+
+	/* Init input tensor */
+	rknn_input inputs[1];
+	memset(inputs, 0, sizeof(inputs));
+	inputs[0].index = 0;
+	inputs[0].type = RKNN_TENSOR_UINT8;
+	inputs[0].size = input_width * input_height * input_channel;
+	inputs[0].fmt = RKNN_TENSOR_NHWC;
+	inputs[0].pass_through = 0;
+
+	/* Init output tensor */
+	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;
+	}
+
+	cv::Mat letter_image;
+	letter_box(input_image, &letter_image, input_width);
+	inputs[0].buf = letter_image.data;
+
+	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);
+
+	// Post process
+	std::vector<float> out_scales;
+	std::vector<uint8_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);
+	}
+
+
+	yolov5_post_process_u8((uint8_t *)outputs[0].buf, (uint8_t *)outputs[1].buf, (uint8_t *)outputs[2].buf, input_height, input_width,
+					   conf_threshold, nms_threshold, out_zps, out_scales, detect_result_group);
+
+
+	/*
+	yolov5_post_process_fp((float *)outputs[0].buf, (float *)outputs[1].buf, (float *)outputs[2].buf, input_height, input_width,
+			            conf_threshold, nms_threshold, &detect_result_group);
+	*/
+
+	rknn_outputs_release(ctx, io_num.n_output, outputs);
+
+	scale_coords(detect_result_group, img_width, img_height, input_width);
+
+	return 0;
+}
+
+int yolov5_detect_release(rknn_context ctx)
+{
+    rknn_destroy(ctx);
+	return 0;
+}
+
+
+std::string base64_encode(unsigned char const* bytes_to_encode, unsigned int in_len) {
+  std::string ret;
+  int i = 0;
+  int j = 0;
+  unsigned char char_array_3[3];
+  unsigned char char_array_4[4];
+
+  while (in_len--) {
+    char_array_3[i++] = *(bytes_to_encode++);
+    if (i == 3) {
+      char_array_4[0] = (char_array_3[0] & 0xfc) >> 2;
+      char_array_4[1] = ((char_array_3[0] & 0x03) << 4) + ((char_array_3[1] & 0xf0) >> 4);
+      char_array_4[2] = ((char_array_3[1] & 0x0f) << 2) + ((char_array_3[2] & 0xc0) >> 6);
+      char_array_4[3] = char_array_3[2] & 0x3f;
+
+      for(i = 0; (i <4) ; i++)
+        ret += base64_chars[char_array_4[i]];
+      i = 0;
+    }
+  }
+
+  if (i) {
+    for(j = i; j < 3; j++)
+      char_array_3[j] = '\0';
+
+    char_array_4[0] = (char_array_3[0] & 0xfc) >> 2;
+    char_array_4[1] = ((char_array_3[0] & 0x03) << 4) + ((char_array_3[1] & 0xf0) >> 4);
+    char_array_4[2] = ((char_array_3[1] & 0x0f) << 2) + ((char_array_3[2] & 0xc0) >> 6);
+    char_array_4[3] = char_array_3[2] & 0x3f;
+
+    for (j = 0; (j < i + 1); j++)
+      ret += base64_chars[char_array_4[j]];
+
+    while((i++ < 3))
+      ret += '=';
+  }
+
+  return ret;
+}
+
+static inline bool is_base64(unsigned char c) {
+  return (isalnum(c) || (c == '+') || (c == '/'));
 }

src/yolov5_detect_postprocess.cpp

@@ -23,6 +23,7 @@
 
 
 static char labels[YOLOV5_CLASS_NUM][30] = {"0", "1"};
+// static char labels[YOLOV5_CLASS_NUM][30] = {"fire"};
 
 const int anchor0[6] = {10, 13, 16, 30, 33, 23};
 const int anchor1[6] = {30, 61, 62, 45, 59, 119};