第三十四章 人脸属性分析实验《DNK210使用指南-SDK版
第三十四章 人脸属性分析实验在上一章节中，介绍了利用KPU模块实现人脸口罩佩戴检测功能，本章将继续介绍利用KPU模块实现人脸属性分析功能。通过本章的学习，读者将学习到使用SDK编程技术实现人脸属性分析应用。本章分为如下几个小节：34.1 KPU模块介绍34.2 硬件设计34.3 程序设计34.4 运行验证
34.1 KPU模块介绍有关KPU模块的介绍，请见第30.1小节《KPU介绍》。34.2 硬件设计34.2.1 例程功能1. 获取摄像头输出的图像，并送入KPU进行人脸检测，接着对检测到的人脸分别进行人脸五关键点检测和属性分析，最后将所有的检测结果和原始图像一同在LCD上进行显示。34.2.2 硬件资源本章实验内容，主要讲解KPU模块的使用，无需关注硬件资源。34.2.3 原理图本章实验内容，主要讲解KPU模块的使用，无需关注原理图。34.3 程序设计34.3.1 main.c代码main.c中的代码如下所示：INCBIN(model_kpu, "face_detect_320x240.kmodel");INCBIN(model_fac, "fac.kmodel");INCBIN(model_ld5, "ld5.kmodel"); image_t kpu_image,crop_image,ai_image; static float g_anchor = {0.1075, 0.126875, 0.126875, 0.175, 0.1465625, 0.2246875, 0.1953125, 0.25375, 0.2440625, 0.351875, 0.341875, 0.4721875, 0.5078125, 0.6696875, 0.8984375, 1.099687, 2.129062, 2.425937}; /*特征检测*/ char *pos_face_attr ={ "Male ", "Mouth Open ", "Smiling ", "Glasses"};char *neg_face_attr ={ "Female ", "Mouth Closed", "No Smile", "No Glasses"}; static volatile uint8_t ai_done_flag; /* KPU运算完成回调 */static void ai_done_callback(void *userdata){ ai_done_flag = 1;} static inline float sigmoid(float x){ return 1.f / (1.f + expf(-x));} int main(void){ uint8_t *disp; uint8_t *ai; kpu_model_context_t task_kpu; kpu_model_context_t task_fac; kpu_model_context_t task_ld5; float *pred_box, *pred_clses,*pred_landm; size_t pred_box_size, pred_clses_size,pred_landm_size; uint16_t x1 = 0, y1 = 0, x2, y2, cut_width, cut_height; uint16_t x_circle = 0; uint16_t y_circle = 0; region_layer_t detect_kpu; obj_info_t cla_obj_coord; sysctl_pll_set_freq(SYSCTL_PLL0, 800000000); sysctl_pll_set_freq(SYSCTL_PLL1, 400000000); sysctl_pll_set_freq(SYSCTL_PLL2, 45158400); sysctl_clock_enable(SYSCTL_CLOCK_AI); sysctl_set_power_mode(SYSCTL_POWER_BANK6, SYSCTL_POWER_V18); sysctl_set_power_mode(SYSCTL_POWER_BANK7, SYSCTL_POWER_V18); sysctl_set_spi0_dvp_data(1); lcd_init(); lcd_set_direction(DIR_YX_LRUD); camera_init(24000000); camera_set_pixformat(PIXFORMAT_RGB565); camera_set_framesize(320, 240); kpu_image.pixel = 3; kpu_image.width = 320; kpu_image.height = 240; // image_init(&kpu_image); ai_image.pixel = 3; ai_image.width = 128; ai_image.height = 128; image_init(&ai_image); if (kpu_load_kmodel(&task_kpu, (const uint8_t *)model_kpu_data) != 0) { printf("Kmodel load failed!\n"); while (1); } if (kpu_load_kmodel(&task_ld5, (const uint8_t *)model_ld5_data) != 0) { printf("Kmodel load failed!\n"); while (1); } if (kpu_load_kmodel(&task_fac, (const uint8_t *)model_fac_data) != 0) { printf("Kmodel load failed!\n"); while (1); } detect_kpu.anchor_number = ANCHOR_NUM; detect_kpu.anchor = g_anchor; detect_kpu.threshold = 0.5; detect_kpu.nms_value = 0.2; region_layer_init(&detect_kpu, 10, 8, 30, 320, 240); while (1) { if (camera_snapshot(&disp, &ai) == 0) { ai_done_flag = 0; if (kpu_run_kmodel(&task_kpu, (const uint8_t *)ai, DMAC_CHANNEL5, ai_done_callback, NULL) != 0) { printf("Kmodel run failed!\n"); while (1); } while (ai_done_flag == 0); if (kpu_get_output(&task_kpu, 0, (uint8_t **)&pred_box, &pred_box_size) != 0) { printf("Output get failed!\n"); while (1); } detect_kpu.input = pred_box; region_layer_run(&detect_kpu, &cla_obj_coord); // region_layer_draw_boxes(&detect_kpu, draw_boxes_callback); for (size_t j = 0; j < cla_obj_coord.obj_number; j++) { if (cla_obj_coord.obj.x1 >= 2) { x1 = cla_obj_coord.obj.x1 - 2; /* 对识别框稍微放大点 */ } if (cla_obj_coord.obj.y1 >= 2) { y1 = cla_obj_coord.obj.y1 - 2; } x2 = cla_obj_coord.obj.x2 + 2; y2 = cla_obj_coord.obj.y2 + 2; draw_box_rgb565_image((uint16_t *)disp, 320, x1, y1, x2, y2, GREEN);/* 画人脸框 */ cut_width = x2 - x1 ; cut_height = y2 - y1 ; kpu_image.addr = ai; crop_image.pixel = 3; crop_image.width = cut_width; crop_image.height = cut_height; image_init(&crop_image); image_crop(&kpu_image, &crop_image, x1, y1); image_resize(&crop_image,&ai_image); image_deinit(&crop_image); ai_done_flag = 0; if (kpu_run_kmodel(&task_ld5, (const uint8_t *)ai_image.addr, DMAC_CHANNEL5, ai_done_callback, NULL) != 0) { printf("Kmodel run failed!\n"); while (1); } while (ai_done_flag == 0); if (kpu_get_output(&task_ld5, 0, (uint8_t **)&pred_landm, &pred_landm_size) != 0) { printf("Output get failed!\n"); while (1); } for (size_t i = 0; i < 5; i++) { x_circle= sigmoid(pred_landm) * cut_width+ x1; y_circle= sigmoid(pred_landm) * cut_height+ y1; draw_point_rgb565_image((uint16_t *)disp, 320, x_circle, y_circle, BLUE);/* 画五个人脸特征点 */ } ai_done_flag = 0; if (kpu_run_kmodel(&task_fac, (const uint8_t *)ai_image.addr, DMAC_CHANNEL5, ai_done_callback, NULL) != 0) { printf("Kmodel run failed!\n"); while (1); } while (ai_done_flag == 0); if (kpu_get_output(&task_fac, 0, (uint8_t **)&pred_clses, &pred_clses_size) != 0) { printf("Output get failed!\n"); while (1); } for (size_t i = 0; i < 4; i++) { if (pred_clses > 0.7) { draw_string_rgb565_image((uint16_t *)disp, 320, 240, x1, y1 + i * 16, pos_face_attr, GREEN); // printf("%s\r\n",pos_face_attr); } else { draw_string_rgb565_image((uint16_t *)disp, 320, 240, x1, y1 + i * 16, neg_face_attr, RED); // printf("%s\r\n",neg_face_attr); } } } lcd_draw_picture(0, 0, 320, 240, (uint16_t *)disp); camera_snapshot_release(); } }}本实验同时使用了三个AI模型，多个模型合理的配合使用，能有效提高识别率。face_detect_320x240.kmodel是人脸检测模型，用于获取人脸关键信息的坐标，然后以图像两个坐标调用image_crop函数把人脸的关键区域切割下来用于其他两个模型的使用。 fac.kmodel是人脸属性分析模型，网络运算的图片大小为128*128。ld5.kmodel是人脸5个关键点检测的模型，网络运算的图片大小为128*128。可以看到一开始是先初始化了LCD和摄像头，初始化完成后创建一个128*128的RGB888图片缓存区，然后加载上述三个需要用到的网络模型，并初始化YOLO2网络，配置region_layer_t结构体参数的数据。最后在一个循环中不断地获取摄像头输出的图像，图像尺寸为320*240，将摄像头图像送入KPU中进行运算，然后将运算结果作为输入传入region_layer_run函数进行解析，该函数会把解析的坐标放在cla_obj_coord结构体中，进而通过两个坐标点提取摄像头图像的人脸区域，我们将人脸区域切割下来，再缩放成128*128的图像大小，最后分别放入人脸表情分析模型和人脸的5个关键点检测模型运算，并将运算结果绘制到LCD显示器上。34.4 运行验证将DNK210开发板连接到电脑主机，通过VSCode将固件烧录到开发板中，将摄像头对准人脸，让其采集到人脸图像，随后便能在LCD上看到摄像头输出的图像，同时能看到图像上标注了人脸位置、人脸五关键点位置、人脸属性等信息，如下图所示： 图34.4.1 LCD显示人脸属性分析实验结果
页:
[1]