Sipeed MAix BiT AIoT 开发套件测评-基础功能测试

zygalaxy

Sipeed MAix BiT AIoT 开发套件测评-基础功能测试 [复制链接]

<div class='showpostmsg'><i class="pstatus"> 本帖最后由 zygalaxy 于 2024-7-12 17:31 编辑 </i>

# K210核心功能验证

## 时间规划
- **评测第三天至第四天**

## 内容要点

### CPU性能测试

- 运行一系列基准测试代码，以评估处理器的运算速度与效率。
这里使用MicroPython开发,所以主要在开发板上运行MicroPython的性能测试代码。

由于时间的限制就从github上面找了一个micropython benchmarks的仓库
这是原作者：[https://github.com/shaoziyang/micropython_benchmarks](https://github.com/shaoziyang/micropython_benchmarks "https://github.com/shaoziyang/micropython_benchmarks")

这是下面的测试的代码：

需要注意的是因为是maixpy的固件，获取时钟频率的方法是
```python
from Maix import freq
try:
  print('System freq: {:.1f} MHz'.format(freq.get()[0]))
except:
  print('System freq: {:.1f} MHz'.format(freq.get_cpu()))
```
注意自行在源代码中更改

```python
import time
import machine
import gc
from Maix import freq

def pi(places=100):
extra = 8
one = 10 ** (places+extra)
t, c, n, na, d, da = 3*one, 3*one, 1, 0, 0, 24

while t > 1: 
  n, na, d, da = n+na, na+8, d+da, da+32
  t = t * n // d
  c += t
return c // (10 ** extra)

def pi_test(n = 5000):
  t1 = time.ticks_ms()
  t = pi(n)
  t2 = time.ticks_ms()
  r = time.ticks_diff(t2, t1)/1000
  print('Pi', n, 'digit calculation: ', r, 's')
  return '%.2f'%r

def int_add_test(n = 1000000, a = 12345, b = 56789):
  t1 = time.ticks_ms()
  sum = 0
  for i in range(n):
  sum = a + b
  t2 = time.ticks_ms()
  r = time.ticks_diff(t2, t1)/1000
  print('Integer Add test', n, 'times: ', r, 's')
  return '%.2f'%r

def float_add_test(n=1000000, a = 1234.5678, b = 5678.1234):
  t1 = time.ticks_ms()
  sum = 0
  for i in range(n):
  sum = a + b
  t2 = time.ticks_ms()
  r = time.ticks_diff(t2, t1)/1000
  print('Float Add test', n, 'times:', r, 's')
  return '%.2f'%r

def int_mul_test(n=1000000, a = 12345, b = 56789):
  t1 = time.ticks_ms()
  sum = 0
  for i in range(n):
  sum = a * b
  t2 = time.ticks_ms()
  r = time.ticks_diff(t2, t1)/1000
  print('Integer Mul test', n, 'times: ', r, 's')
  return '%.2f'%r

def float_mul_test(n=1000000, a = 1234.5678, b = 5678.1234):
  t1 = time.ticks_ms()
  sum = 0
  for i in range(n):
  sum = a * b
  t2 = time.ticks_ms()
  r = time.ticks_diff(t2, t1)/1000
  print('Float Mul test', n, 'times: ', r, 's')
  return '%.2f'%r

def int_div_test(n=1000000, a = 123456, b = 567):
  t1 = time.ticks_ms()
  sum = 0
  for i in range(n):
  sum = a // b
  t2 = time.ticks_ms()
  r = time.ticks_diff(t2, t1)/1000
  print('Integer Div test', n, 'times: ', r, 's')
  return '%.2f'%r

def float_div_test(n=1000000, a = 12345.678, b = 56.789):
  t1 = time.ticks_ms()
  sum = 0
  for i in range(n):
  sum = a / b
  t2 = time.ticks_ms()
  r = time.ticks_diff(t2, t1)/1000
  print('Float Div test', n, 'times: ', r, 's')
  return '%.2f'%r

def mem():
  r = gc.mem_free()
  print('free memory:', r)

print('Speed test')
try:
  print('System freq: {:.1f} MHz'.format(freq.get()[0]))
except:
  print('System freq: {:.1f} MHz'.format(freq.get_cpu()))
    
print('\nCalcaulate integer addition')
gc.collect()
mem()
d1 = int_add_test()
d2 = int_add_test()
d3 = int_add_test()
r_int_add =min(d1, d2, d3)
print('Integer addition test result: ', r_int_add, 's')
mem()

print('\nCalcaulate float addition')
gc.collect()
mem()
d1 = float_add_test()
d2 = float_add_test()
d3 = float_add_test()
r_float_add = min(d1, d2, d3)
print('Float addition test result: ', r_float_add, 's')
mem()

print('\nCalcaulate integer multiplication')
gc.collect()
mem()
d1 = int_mul_test()
d2 = int_mul_test()
d3 = int_mul_test()
r_int_mul = min(d1, d2, d3)
print('Integer multiplication test result: ', r_int_mul, 's')
mem()

print('\nCalcaulate float multiplication')
gc.collect()
mem()
d1 = float_mul_test()
d2 = float_mul_test()
d3 = float_mul_test()
r_float_mul = min(d1, d2, d3)
print('Float multiplication test result: ', r_float_mul, 's')
mem()

print('\nCalcaulate integer division')
gc.collect()
mem()
d1 = int_div_test()
d2 = int_div_test()
d3 = int_div_test()
r_int_div = min(d1, d2, d3)
print('Integer division test result: ', r_int_div, 's')
mem()

print('\nCalcaulate float division')
gc.collect()
mem()
d1 = float_div_test()
d2 = float_div_test()
d3 = float_div_test()
r_float_div = min(d1, d2, d3)
print('Float division test result: ', r_float_div, 's')
mem()

print('\nCalcaulate Pi 1000 digit')
gc.collect()
mem()
try:
  d1 = pi_test(1000)
  d2 = pi_test(1000)
  d3 = pi_test(1000)
  r_pi_1000 = min(d1, d2, d3)
  print('1000 digit Pi calculation result: ', r_pi_1000, 's')
  mem()
except:
  r_pi_1000 = None
  print('calculation error')

print('Test result:')
print('Integer addition test result: ', r_int_add, 's')
print('Float addition test result: ', r_float_add, 's')
print('Integer multiplication test result: ', r_int_mul, 's')
print('Float multiplication test result: ', r_float_mul, 's')
print('Integer division test result: ', r_int_div, 's')
print('Float division test result: ', r_float_div, 's')
print('1000 digit Pi calculation result: ', r_pi_1000, 's')
```
跑完的结果如下：

通过数据能看出来CPU的性能是不错的

```python
>>> Speed test
System freq: 403.0 MHz

Calcaulate integer addition
free memory: 508064
Integer Add test 1000000 times:1.33 s
Integer Add test 1000000 times:1.331 s
Integer Add test 1000000 times:1.329 s
Integer addition test result:1.33 s
free memory: 507680

Calcaulate float addition
free memory: 507776
Float Add test 1000000 times: 2.616 s
Float Add test 1000000 times: 2.623 s
Float Add test 1000000 times: 2.62 s
Float addition test result:2.62 s
free memory: 411136

Calcaulate integer multiplication
free memory: 507616
Integer Mul test 1000000 times:1.522 s
Integer Mul test 1000000 times:1.504 s
Integer Mul test 1000000 times:1.521 s
Integer multiplication test result:1.50 s
free memory: 507328

Calcaulate float multiplication
free memory: 507552
Float Mul test 1000000 times:2.637 s
Float Mul test 1000000 times:2.634 s
Float Mul test 1000000 times:2.641 s
Float multiplication test result:2.63 s
free memory: 390848

Calcaulate integer division
free memory: 507488
Integer Div test 1000000 times:1.421 s
Integer Div test 1000000 times:1.37 s
Integer Div test 1000000 times:1.371 s
Integer division test result:1.37 s
free memory: 507200

Calcaulate float division
free memory: 507424
Float Div test 1000000 times:2.696 s
Float Div test 1000000 times:2.7 s
Float Div test 1000000 times:2.697 s
Float division test result:2.70 s
free memory: 366688

Calcaulate Pi 1000 digit
free memory: 507360
Pi 1000 digit calculation:0.126 s
Pi 1000 digit calculation:0.125 s
Pi 1000 digit calculation:0.122 s
1000 digit Pi calculation result:0.12 s
free memory: 218880
Test result:
Integer addition test result:1.33 s
Float addition test result:2.62 s
Integer multiplication test result:1.50 s
Float multiplication test result:2.63 s
Integer division test result:1.37 s
Float division test result:2.70 s
1000 digit Pi calculation result:0.12 s
```

### KPU应用尝试

- 利用预置模型进行图像识别或简单物体检测，评估AI加速器的性能。

这里我们去[maixhub](https://maixhub.com/ "maixhub") 找一个人脸检测的模型进行测试 -人脸检测模型https://maixhub.com/model/zoo/68

解压下载后的的压缩包可以找到要运行的模型和测试模型的代码

### 文末可以下载

<ignore_js_op>

<div class="tip tip_4 aimg_tip" id="aimg_823517_menu" style="position: absolute; display: none" disautofocus="true">
<div class="xs0">
<p><strong>Snipaste_2024-07-12_16-47-49.png</strong> <em class="xg1">(5.94 KB, 下载次数: 0)</em></p>
<p>
<a href="javascript:;" class="loginf" >下载附件</a>

<a href="javascript:;" onclick="showWindow(this.id, this.getAttribute('url'), 'get', 0);" id="savephoto_823517" url="home.php?mod=spacecp&ac=album&op=saveforumphoto&aid=823517&handlekey=savephoto_823517">保存到相册</a>

</p>

</div>
<div class="tip_horn"></div>
</div>

</ignore_js_op>

模型我们可以放在flash中读取并运行也可以放在SD卡中进行读取并运行，这里为了检测就直接使用`kflash`工具刷入指定的flash地址。

<ignore_js_op>

<div class="tip tip_4 aimg_tip" id="aimg_823531_menu" style="position: absolute; display: none" disautofocus="true">
<div class="xs0">
<p><strong>Snipaste_2024-07-12_16-54-20.png</strong> <em class="xg1">(17.48 KB, 下载次数: 0)</em></p>
<p>
<a href="javascript:;" class="loginf" >下载附件</a>

<a href="javascript:;" onclick="showWindow(this.id, this.getAttribute('url'), 'get', 0);" id="savephoto_823531" url="home.php?mod=spacecp&ac=album&op=saveforumphoto&aid=823531&handlekey=savephoto_823531">保存到相册</a>

</p>

</div>
<div class="tip_horn"></div>
</div>

</ignore_js_op>

然后将代码放入IDE进行运行

注意这里的屏幕可能会出现反转和镜像，修改如下代码为
```
    # sensor.set_hmirror(sensor_hmirror)
  # sensor.set_vflip(sensor_vflip)
    sensor.set_hmirror(True)
  sensor.set_vflip(True)
    ```

```python

import sensor, image, lcd, time
import KPU as kpu
import gc, sys

def lcd_show_except(e):
  import uio
  err_str = uio.StringIO()
  sys.print_exception(e, err_str)
  err_str = err_str.getvalue()
  img = image.Image(size=(224,224))
  img.draw_string(0, 10, err_str, scale=1, color=(0xff,0x00,0x00))
  lcd.display(img)

def main(model_addr=0x300000, lcd_rotation=0, sensor_hmirror=False, sensor_vflip=False):
  try:
  sensor.reset()
  except Exception as e:
  raise Exception("sensor reset fail, please check hardware connection, or hardware damaged! err: {}".format(e))
  sensor.set_pixformat(sensor.RGB565)
  sensor.set_framesize(sensor.QVGA)
  sensor.set_hmirror(sensor_hmirror)
  sensor.set_vflip(sensor_vflip)
  sensor.run(1)

lcd.init(type=1)
  lcd.rotation(lcd_rotation)
  lcd.clear(lcd.WHITE)

anchors = (1.889, 2.5245, 2.9465, 3.94056, 3.99987, 5.3658, 5.155437, 6.92275, 6.718375, 9.01025)
  try:
  task = None
  task = kpu.load(model_addr)
  kpu.init_yolo2(task, 0.5, 0.3, 5, anchors) # threshold:[0,1], nms_value: [0, 1]
  while(True):
    img = sensor.snapshot()
    t = time.ticks_ms()
    objects = kpu.run_yolo2(task, img)
    t = time.ticks_ms() - t
    if objects:
      for obj in objects:
      img.draw_rectangle(obj.rect())
    img.draw_string(0, 200, "t:%dms" %(t), scale=2)
    lcd.display(img)
  except Exception as e:
  raise e
  finally:
  if not task is None:
    kpu.deinit(task)

if __name__ == "__main__":
  try:
  main( model_addr=0x300000, lcd_rotation=0, sensor_hmirror=False, sensor_vflip=False)
  # main(model_addr="/sd/m.kmodel")
  except Exception as e:
  sys.print_exception(e)
  lcd_show_except(e)
  finally:
  gc.collect()

```
此时我们寻找一张全家福查看识别的结果

<ignore_js_op>

<a href="javascript:;" onclick="showWindow(this.id, this.getAttribute('url'), 'get', 0);" id="savephoto_823532" url="home.php?mod=spacecp&ac=album&op=saveforumphoto&aid=823532&handlekey=savephoto_823532">保存到相册</a>

</p>

</div>
<div class="tip_horn"></div>
</div>

</ignore_js_op>

<ignore_js_op>

<a href="javascript:;" onclick="showWindow(this.id, this.getAttribute('url'), 'get', 0);" id="savephoto_823533" url="home.php?mod=spacecp&ac=album&op=saveforumphoto&aid=823533&handlekey=savephoto_823533">保存到相册</a>

</p>

</div>
<div class="tip_horn"></div>
</div>

</ignore_js_op>

我们可以看到单人人脸检测的帧率还是非常高的，而且在多人的情况下，帧率也在一个比较理想的范围，由此可见其kpu算力的强大。

- k210还有一些其他比较强大的功能，如色块的检测

```python
import sensor
import image
import lcd
import time
lcd.init()
sensor.reset()
sensor.set_pixformat(sensor.RGB565)
sensor.set_framesize(sensor.QVGA)
sensor.run(1)
green_threshold = (0, 80,-70, -10, -0, 30)
while True:
  img=sensor.snapshot()
  blobs = img.find_blobs([green_threshold])
  if blobs:
  for b in blobs:
    tmp=img.draw_rectangle(b[0:4])
    tmp=img.draw_cross(b[5], b[6])
    c=img.get_pixel(b[5], b[6])
  lcd.display(img)

```

### I/O与外设交互

- 测试GPIO控制、SPI、I2C等接口功能，确保与外部设备通信无碍。

目前手头只有I2C的光照强度传感器BH1750，所以以为基础来测试k210的I2C功能。

如下代码可以查看当前光照强度：

```python
from machine import I2C
import time

# 初始化I2C
i2c = I2C(I2C.I2C0, freq=100000, scl=25, sda=24)

def read_bh1750():
  # 发送测量命令
  i2c.writeto(0x23, bytearray([0x10]))
  time.sleep(0.018) # 等待测量完成
  # 读取数据
  data = i2c.readfrom(0x23, 2)
  lux = ((data[0] << 8) | data[1]) / 1.2
  return lux

# 持续读取光照强度
while True:
  lux = read_bh1750()
  print("Lux:", lux)
  time.sleep(1) # 每秒读取一次，可以根据需要调整时间间隔

```
可以看到已经能够读取当前的光照强度了

<ignore_js_op>

<div class="tip tip_4 aimg_tip" id="aimg_823536_menu" style="position: absolute; display: none" disautofocus="true">
<div class="xs0">
<p><strong>Snipaste_2024-07-12_17-28-03.png</strong> <em class="xg1">(5.95 KB, 下载次数: 0)</em></p>
<p>
<a href="javascript:;" class="loginf" >下载附件</a>

<a href="javascript:;" onclick="showWindow(this.id, this.getAttribute('url'), 'get', 0);" id="savephoto_823536" url="home.php?mod=spacecp&ac=album&op=saveforumphoto&aid=823536&handlekey=savephoto_823536">保存到相册</a>

</p>

</div>
<div class="tip_horn"></div>
</div>

</ignore_js_op>

光照强度的检测在识别中还是很重要的一个因素，当环境光较暗的时候可以通过补光来达到更好的识别效果。

其他的接口与功能会在后续的项目中进行使用~
---</div><script>					var loginstr = '<div class="locked">查看本帖全部内容，请<a href="javascript:;"   style="color:#e60000" class="loginf">登录</a>或者<a href="https://bbs.eeworld.com.cn/member.php?mod=register_eeworld.php&action=wechat" style="color:#e60000" target="_blank">注册</a></div>';
					
					if(parseInt(discuz_uid)==0){
						
					}		</script><script type="text/javascript">(function(d,c){var a=d.createElement("script"),m=d.getElementsByTagName("script")[0],eewurl="//counter.eeworld.com.cn/pv/count/";a.src=eewurl+c;m.parentNode.insertBefore(a,m)})(document,523)</script>

Jacktang

光照强度的检测在识别中还是很重要的一个因素，当环境光较暗的时候可以通过补光来达到更好的识别效果。

这就是补光的意义

Sipeed MAix BiT AIoT 开发套件测评-基础功能测试 [复制链接]

最新回复

浏览过的版块