期待结果和实际结果
能发送出去我的信息即可
错误日志
有图
尝试解决过程
改变数据格式
补充材料
2 Answers
from libs.PipeLine import PipeLine, ScopedTiming
from libs.AIBase import AIBase
from libs.AI2D import Ai2d
import os
import ujson
from media.media import *
from time import *
from machine import UART
from machine import FPIOA
import nncase_runtime as nn
import ulab.numpy as np
import time
import utime
import image
import random
import gc
import sys
import aicube
flag = 0
final_num = 0
change_flag = 0
left_flag = 0
right_flag = 0
go_straight_falg = 0
two_flag = 0
four_flag = 0
#num_list = []
#length = 10
#for _ in range(length):
# my_list.append(None)
# UART2代码
fpioa = FPIOA()
fpioa.set_function(11,FPIOA.UART2_TXD)
fpioa.set_function(12,FPIOA.UART2_RXD)
uart=UART(UART.UART2,115200) #设置串口号2和波特率
class STM32_transmit(): # 定义 STM32 发送类
head1 = 0x2C # uint8_t 帧头
head2 = 0x12 # uint8_t 帧头
flag1 = 0 # uint8_t 标志位
flag2 = 0 # uint8_t 标志位
flag3 = 0 # uint8_t 标志位
flag4 = 0 # uint8_t 标志位
flag5 = 0 # uint8_t 标志位
flag6 = 0 # uint8_t 标志位
flag7 = 0 # uint8_t 标志位
flag8 = 0 # uint8_t 标志位
tail = 0x5B # uint8_t 帧尾
# 实例化类
TSTM32 = STM32_transmit() # 实例化 STM32_transmit() 为 TSTM32
# 定义打包函数
def TSTM32_data(): # 数据打包函数
data=bytearray([TSTM32.head1, # 帧头
TSTM32.head2, # 帧头
TSTM32.flag1, # 标志位
TSTM32.flag2, # 标志位
TSTM32.flag3, # 标志位
TSTM32.flag4, # 标志位
TSTM32.flag5, # 标志位
TSTM32.flag6, # 标志位
TSTM32.flag7, # 标志位
TSTM32.flag8, # 标志位
TSTM32.tail # 帧尾
])
return data # 返回打包好的数据
#uart1.write(TSTM32_data())
# 自定义目标检测任务类
class DetectionApp(AIBase):
def __init__(self,kmodel_path,labels,model_input_size=[640,640],anchors=[10.13,16,30,33,23,30,61,62,45,59,119,116,90,156,198,373,326],model_type="AnchorBaseDet",confidence_threshold=0.5,nms_threshold=0.25,nms_option=False,strides=[8,16,32],rgb888p_size=[1280,720],display_size=[1920,1080],debug_mode=0):
super().__init__(kmodel_path,model_input_size,rgb888p_size,debug_mode)
# kmodel路径
self.kmodel_path=kmodel_path
# 类别标签
self.labels=labels
# 模型输入分辨率
self.model_input_size=model_input_size
# 检测任务的锚框
self.anchors=anchors
# 模型类型，支持"AnchorBaseDet","AnchorFreeDet","GFLDet"三种模型
self.model_type=model_type
# 检测框类别置信度阈值
self.confidence_threshold=confidence_threshold
# 检测框NMS筛选阈值
self.nms_threshold=nms_threshold
# NMS选项，如果为True做类间NMS,如果为False做类内NMS
self.nms_option=nms_option
# 输出特征图的降采样倍数
self.strides=strides
# sensor给到AI的图像分辨率，宽16字节对齐
self.rgb888p_size=[ALIGN_UP(rgb888p_size[0],16),rgb888p_size[1]]
# 视频输出VO分辨率，宽16字节对齐
self.display_size=[ALIGN_UP(display_size[0],16),display_size[1]]
# 调试模式
self.debug_mode=debug_mode
# 检测框预置颜色值
self.color_four=[(255, 220, 20, 60), (255, 119, 11, 32), (255, 0, 0, 142), (255, 0, 0, 230),
(255, 106, 0, 228), (255, 0, 60, 100), (255, 0, 80, 100), (255, 0, 0, 70),
(255, 0, 0, 192), (255, 250, 170, 30), (255, 100, 170, 30), (255, 220, 220, 0),
(255, 175, 116, 175), (255, 250, 0, 30), (255, 165, 42, 42), (255, 255, 77, 255),
(255, 0, 226, 252), (255, 182, 182, 255), (255, 0, 82, 0), (255, 120, 166, 157)]
# Ai2d实例，用于实现模型预处理
self.ai2d=Ai2d(debug_mode)
# 设置Ai2d的输入输出格式和类型
self.ai2d.set_ai2d_dtype(nn.ai2d_format.NCHW_FMT,nn.ai2d_format.NCHW_FMT,np.uint8, np.uint8)
# 配置预处理操作，这里使用了pad和resize，Ai2d支持crop/shift/pad/resize/affine，具体代码请打开/sdcard/app/libs/AI2D.py查看
def config_preprocess(self,input_image_size=None):
with ScopedTiming("set preprocess config",self.debug_mode > 0):
# 初始化ai2d预处理配置，默认为sensor给到AI的尺寸，您可以通过设置input_image_size自行修改输入尺寸
ai2d_input_size=input_image_size if input_image_size else self.rgb888p_size
# 计算padding参数
top,bottom,left,right=self.get_padding_param()
# 配置padding预处理
self.ai2d.pad([0,0,0,0,top,bottom,left,right], 0, [114,114,114])
# 配置resize预处理
self.ai2d.resize(nn.interp_method.tf_bilinear, nn.interp_mode.half_pixel)
# build预处理过程，参数为输入tensor的shape和输出tensor的shape
self.ai2d.build([1,3,ai2d_input_size[1],ai2d_input_size[0]],[1,3,self.model_input_size[1],self.model_input_size[0]])
# 自定义当前任务的后处理,这里调用了aicube模块的后处理接口
def postprocess(self,results):
with ScopedTiming("postprocess",self.debug_mode > 0):
# AnchorBaseDet模型的后处理
if self.model_type == "AnchorBaseDet":
det_boxes = aicube.anchorbasedet_post_process( results[0], results[1], results[2], self.model_input_size, self.rgb888p_size, self.strides, len(labels), self.confidence_threshold, self.nms_threshold, self.anchors, self.nms_option)
# GFLDet模型的后处理
elif self.model_type == "GFLDet":
det_boxes = aicube.gfldet_post_process( results[0], results[1], results[2], self.model_input_size, self.rgb888p_size, self.strides, len(labels), self.confidence_threshold, self.nms_threshold, self.nms_option)
# AnchorFreeDet模型的后处理
elif self.model_type=="AnchorFreeDet":
det_boxes = aicube.anchorfreedet_post_process( results[0], results[1], results[2], self.model_input_size, self.rgb888p_size, self.strides, len(labels), self.confidence_threshold, self.nms_threshold, self.nms_option)
else:
det_boxes=None
return det_boxes
# 将结果绘制到屏幕上
def draw_result(self,pl,det_boxes):
global change_flag
global final_num
global left_flag
global right_flag
global go_straight_falg
global two_flag
global four_flag
with ScopedTiming("draw osd",self.debug_mode > 0):
if det_boxes:
#time.sleep(0.3)
pl.osd_img.clear()
for det_boxe in det_boxes:
# 获取每一个检测框的坐标，并将其从原图分辨率坐标转换到屏幕分辨率坐标，将框和类别信息绘制在屏幕上
if len(det_boxes) == 1 and change_flag == 0:
#找框，找出参数
#---------------------------------------------------------------------------------------------------------------------------------
x1, y1, x2, y2 = det_boxe[2],det_boxe[3],det_boxe[4],det_boxe[5]
sx=int(x1 * self.display_size[0] // self.rgb888p_size[0])
sy=int(y1 * self.display_size[1] // self.rgb888p_size[1])
w = int(float(x2 - x1) * self.display_size[0] // self.rgb888p_size[0])
h = int(float(y2 - y1) * self.display_size[1] // self.rgb888p_size[1])
pl.osd_img.draw_rectangle(sx , sy , w , h , color=self.get_color(det_boxe[0]) ,thickness = 4)
label = self.labels[det_boxe[0]]
score = str(round(det_boxe[1],2))
pl.osd_img.draw_string_advanced(sx, sy-50,32, label + " " + score , color=self.get_color(det_boxe[0]))
#---------------------------------------------------------------------------------------------------------------------------------
# print(sx)
# print(sy)
final_num = int(label)#存目标房间号
change_flag += 1 #改变标志位以防止误识别进入该判断
TSTM32.flag1 = final_num
uart.write(TSTM32_data())
if len(det_boxes) == 2:#确定是两个数字
# print(final_num)
# pl.osd_img.clear()
# time.sleep(0.3)
# if two_flag == 0 :#防止误识别
## two_flag += 1
#找框，找出参数
#--------------------------------------------------------------------------------------------------------------------------------
x1, y1, x2, y2 = det_boxe[2],det_boxe[3],det_boxe[4],det_boxe[5]
sx=int(x1 * self.display_size[0] // self.rgb888p_size[0])
sy=int(y1 * self.display_size[1] // self.rgb888p_size[1])
w = int(float(x2 - x1) * self.display_size[0] // self.rgb888p_size[0])
h = int(float(y2 - y1) * self.display_size[1] // self.rgb888p_size[1])
pl.osd_img.draw_rectangle(sx , sy , w , h , color=self.get_color(det_boxe[0]) ,thickness = 4)
label = self.labels[det_boxe[0]]
score = str(round(det_boxe[1],2))
pl.osd_img.draw_string_advanced(sx, sy-50,32, label + " " + score , color=self.get_color(det_boxe[0]))
#---------------------------------------------------------------------------------------------------------------------------------
#数据清理以及判断方向
if int(label) == final_num:
if sx < 300:
left_flag = 1
TSTM32.flag1 = left_flag
uart.write(TSTM32_data())
# uart.write(bytearray(left_flag))
left_flag = 0
else:
right_flag = 2
# print("右走")
TSTM32.flag1 = right_flag
uart.write(TSTM32_data())
# uart.write(bytearray(right_flag))
right_flag = 0
if len(det_boxes) == 4 :
#找框，找出参数
#---------------------------------------------------------------------------------------------------------------------------
x1, y1, x2, y2 = det_boxe[2],det_boxe[3],det_boxe[4],det_boxe[5]
sx=int(x1 * self.display_size[0] // self.rgb888p_size[0])
sy=int(y1 * self.display_size[1] // self.rgb888p_size[1])
w = int(float(x2 - x1) * self.display_size[0] // self.rgb888p_size[0])
h = int(float(y2 - y1) * self.display_size[1] // self.rgb888p_size[1])
pl.osd_img.draw_rectangle(sx , sy , w , h , color=self.get_color(det_boxe[0]) ,thickness = 4)
label = self.labels[det_boxe[0]]
score = str(round(det_boxe[1],2))
pl.osd_img.draw_string_advanced(sx, sy-50,32, label + " " + score , color=self.get_color(det_boxe[0]))
#----------------------------------------------------------------------------------------------------------------------------
#数据清理以及判断方向
if int(label) == final_num:
if sx < 300:
left_flag = 1
# print("左走")
TSTM32.flag1 = left_flag
uart.write(TSTM32_data())
# uart.write(bytearray(left_flag))
left_flag = 0
else:
right_flag = 2
# print("右走")
TSTM32.flag1 = left_flag
uart.write(TSTM32_data())
# uart.write(bytearray(right_flag))
right_flag = 0
else:
pl.osd_img.clear()
pl.osd_img.draw_rectangle(0, 0, 128, 128, color=(0,0,0,0))
# 计算padding参数
def get_padding_param(self):
ratiow = float(self.model_input_size[0]) / self.rgb888p_size[0];
ratioh = float(self.model_input_size[1]) / self.rgb888p_size[1];
ratio = min(ratiow, ratioh)
new_w = int(ratio * self.rgb888p_size[0])
new_h = int(ratio * self.rgb888p_size[1])
dw = float(self.model_input_size[0]- new_w) / 2
dh = float(self.model_input_size[1] - new_h) / 2
top = int(round(dh - 0.1))
bottom = int(round(dh + 0.1))
left = int(round(dw - 0.1))
right = int(round(dw - 0.1))
return top,bottom,left,right
# 根据当前类别索引获取框的颜色
def get_color(self, x):
idx=x%len(self.color_four)
return self.color_four[idx]
if __name__=="__main__":
# 添加显示模式，支持"hdmi"和"lcd"
display_mode="lcd"
if display_mode=="lcd":
display_size=[800,480]
else:
display_size=[800,480]
# kmodel路径
kmodel_path="/sdcard/app/tests/ai_test_kmodel/best_AnchorBaseDet_can2_5_n_20241120121751.kmodel"
# 检测类别标签
labels=[
"0",
"9",
"1",
"2",
"3",
"4",
"5",
"6",
"7",
"8"
]
# 类别置信度阈值
confidence_threshold=0.5
# nms阈值
nms_threshold = 0.5
# 训练中使用的锚框，在线训练平台和AICube部署包的deploy_config.json文件中包含该字段，只有AnchorBaseDet需要该参数
anchors=[
104,
200,
140,
151,
121,
181
,
147,
180,
136,
212,
162,
207
,
179,
223,
202,
261,
265,
352
]
# 初始化PipeLine，只关注传给AI的图像分辨率，显示的分辨率
pl=PipeLine(rgb888p_size=[1280,720],display_size=display_size,display_mode=display_mode)
pl.create()
# 检测类实例，关注模型输入分辨率，传给AI的图像分辨率，显示的分辨率
det=DetectionApp(kmodel_path,labels,model_input_size=[640,640],anchors=anchors,rgb888p_size=[1280,720],display_size=display_size,debug_mode=0)
# 配置预处理过程
det.config_preprocess()
try:
while True:
os.exitpoint()
with ScopedTiming("total",1):
# 获取当前帧
img=pl.get_frame()
# 获得检测框
det_boxes=det.run(img)
# 绘制检测框和类别信息
det.draw_result(pl,det_boxes)
# 显示当前的绘制结果
pl.show_image()
gc.collect()
except BaseException as e:
sys.print_exception(e)
finally:
det.deinit()
pl.destroy()
这是我的源码
