python-help/md/ai.md

人工智慧

自訂模型識別

from AI import CustomModel

模型初始化

CustomModel(labels,path)

參數說明

參數	值	描述
labels	列表	標籤列表，必填
path	字串	模型路徑，必填

載入識別

loadModel(img)

參數說明

參數	值	描述
img	圖片	要識別的Image對象，必填

樣例

from cocopiPython import cameraPython,Screen
from AI import CustomModel

Screen = Screen() # 畫布初始化
camera = cameraPython() # 攝像頭初始化

labels = ["load", "unload", "left", "right", "forward", "stop"] # 標籤
path = "/root/preset/model/autologistics" # 模型路徑
CustomModel = CustomModel(labels,path) # 初始化模型

while True:
    canvas = camera.capture() # 獲取攝像頭圖像
    canvas = canvas.crop(48,8,224,224) # 裁剪圖像為224*224
    data = CustomModel.load(canvas) # 載入模型進行識別
    for i in data:
        canvas.draw_rectangle(i["X"],i["Y"],i["X"]+i["W"],i["Y"]+i["H"],color=(255,0,0),thickness=1) # 在螢幕繪製檢測框
        canvas.draw_string(i["X"],i["Y"]-20,(str(i["name"])+":"+str(i["score"])),color=(255,0,0)) # 在螢幕上繪製名稱與置信度
    Screen.show(canvas,48,8) # 顯示畫布

其中i["X"]表示識別結果的檢測框起點X軸坐標，i["Y"]表示識別結果的檢測框起點Y軸坐標，i["W"]表示識別結果的檢測框寬度，i["H"]表示識別結果的檢測框高度，i["name"]表示識別結果的名稱，i["score"]表示識別結果的置信度，i["middleX"]表示識別結果的檢測框中心點的X軸坐標，i["middleY"]表示識別結果的檢測框中心點的Y軸坐標。

邊緣檢測

from AI import EdgeDetection

模型初始化

EdgeDetection()

載入識別

loadModel(img)

參數說明

參數	值	描述
img	圖片	要識別的Image對象，必填

樣例

from cocopiPython import cameraPython,Screen
from AI import EdgeDetection

Screen = Screen() # 畫布初始化
camera = cameraPython() # 攝像頭初始化
EdgeDetection = EdgeDetection() # 初始化

while True:
    img = camera.capture() # 獲取攝像頭資料
    img = img.crop(48,8,224,224) # 裁剪尺寸為224*224
    canvas = EdgeDetection.load(img) # 載入模型
    Screen.show(canvas,48,8) # 顯示畫布

人臉檢測

from AI import Facedetection

模型初始化

Facedetection()

載入識別

loadModel(img)

參數說明

參數	值	描述
img	圖片	要識別的Image對象，必填

樣例

from cocopiPython import cameraPython,Screen
from AI import Facedetection

Screen = Screen() # 畫布初始化
camera = cameraPython() # 攝像頭初始化
face = Facedetection() # 人臉檢測初始化

while True:
    canvas = camera.capture() # 獲取攝像頭圖像
    canvas = canvas.crop(48,8,224,224) # 裁剪圖像尺寸
    data = face.load(canvas) # 載入人臉檢測模型
    for i in data:
        canvas.draw_rectangle(i["X"],i["Y"],i["X"]+i["W"],i["Y"]+i["H"],color=(255,0,0), thickness=1) # 在螢幕上繪製人臉檢測框
        canvas.draw_string(5,5,str(len(data)), scale = 1, color = (255,255,255) , thickness = 1) # 在螢幕上顯示人臉數量
    Screen.show(canvas,48,8) # 顯示畫布

人臉識別

from AI import FaceRecognition

模型初始化

FaceRecognition()

載入模型

載入人臉檢測模型，獲取人臉資訊

loadmodel(img)

參數說明

參數	值	描述
img	圖片	要識別的Image對象，必填

添加人臉

錄入人臉資訊保存到本地

add(name,feature,dataPath)

參數說明

參數	值	描述
name	字串	識別到的人臉名稱，必填
feature	列表	人臉資料，必填
dataPath	字串	保存路徑，選填，預設為"/root/user/model/recorded_face_features.py"

刪除人臉

刪除本地指定人臉資料

remove(name,dataPath)

參數說明

參數	值	描述
name	字串	要刪除的人臉名稱，必填
dataPath	字串	保存路徑，選填，預設為"/root/user/model/recorded_face_features.py"

載入人臉信息

載入人臉採集時保存到本地的人臉資訊資料

loadface(DataPath)

參數說明

參數	值	描述
dataPath	字串	保存路徑，選填，預設為""/root/user/model/recorded_face_features.py"

樣例

人臉採集

import time
from cocopiPython import cameraPython,Screen
from AI import FaceRecognition
from CocoPi import BUTTON

# 按鈕初始化
key_C = BUTTON(13)
key_D = BUTTON(7)

Screen = Screen()
camera = cameraPython()

FaceRecognition = FaceRecognition()
People_list = ["小明","小紅","小剛","小花"]

FaceNum = 0
faceScore = 0
color = (255,0,0)

while True:
    canvas = camera.capture() # 獲取攝像頭資料
    canvas = canvas.crop(48,8,224,224) # 裁剪圖片尺寸
    data = FaceRecognition.loadmodel(canvas) # 載入人臉檢測模型
    for i in data:
        canvas.draw_rectangle(i["X"],i["Y"],i["X"]+i["W"],i["Y"]+i["H"],color=color,thickness=1) # 檢測框
        if key_C.is_pressed() and FaceNum < len(People_list):
            while not key_C.is_pressed()== False:
                time.sleep(0.01)
            FaceRecognition.add(People_list[len(FaceRecognition)],i["feature"]) # 錄入人臉保存
            canvas.draw_rectangle(0,0,320,20,color=(0,0,0),thickness=-1)
            canvas.draw_string(0,0,"已採集"+str(People_list[FaceNum])+"人臉數據",scale=1,color=(255,255,255),thickness=1)
            Screen.show(canvas,48,8)
            FaceNum = FaceNum + 1
            time.sleep(1)
    if key_D.is_pressed() and FaceNum > 0:
        while not key_D.is_pressed()== False:
            time.sleep(0.01)
        FaceRecognition.remove(People_list[len(FaceRecognition)-1]) # 刪除人臉
        canvas.draw_rectangle(0,0,320,20,color=(0,0,0),thickness=-1)
        canvas.draw_string(0,0,"已刪除"+str(People_list[FaceNum-1])+"人臉數據",scale=1,color=(255,255,255),thickness=1)
        Screen.show(canvas,48,8)
        FaceNum = FaceNum - 1
        time.sleep(1)
    Screen.show(canvas,48,8)

人臉識別

from cocopiPython import cameraPython,Screen
from AI import FaceRecognition

Screen = Screen()
camera = cameraPython()
FaceRecognition = FaceRecognition()

faceScore = 0
color = (255,0,0)
    
while True:
    canvas = camera.capture() # 獲取攝像頭資料
    canvas = canvas.crop(48,8,224,224) # 裁剪圖片尺寸
    data = FaceRecognition.loadmodel(canvas) # 載入人臉檢測模型
    FaceRecognition.loadface()
    for i in data:
        faceScore = round(i["recognizeScore"],2)
        canvas.draw_rectangle(i["X"],i["Y"],i["X"]+i["W"],i["Y"]+i["H"],color=color,thickness=1) # 檢測框
        canvas.draw_string(i["X"],i["Y"],str(i["name"])+":"+str(faceScore),scale=1,color=color,thickness=1)
        if faceScore > 80:
            color = (0,255,0)
        else:
            color = (255,0,0)
    Screen.show(canvas,48,8)

人臉採集並識別

import time
from cocopiPython import cameraPython,Screen
from AI import FaceRecognition
from CocoPi import BUTTON

# 按鈕初始化
key_C = BUTTON(13)
key_D = BUTTON(7)

Screen = Screen()
camera = cameraPython()

FaceRecognition = FaceRecognition()
People_list = ["小明","小紅","小剛","小花"]

FaceNum = 0
faceScore = 0
color = (255,0,0)

while True:
    canvas = camera.capture() # 獲取攝像頭圖像
    canvas = canvas.crop(48,8,224,224) # 裁剪圖片尺寸
    data = FaceRecognition.loadmodel(canvas) # 載入人臉檢測模型
    for i in data:
        canvas.draw_rectangle(i["X"],i["Y"],i["X"]+i["W"],i["Y"]+i["H"],color=color,thickness=1) # 檢測框
        if key_C.is_pressed() and FaceNum < len(People_list):
            while not key_C.is_pressed()== False:
                time.sleep(0.01)
            FaceRecognition.add(People_list[len(FaceRecognition)],i["feature"]) # 錄入人臉保存
            canvas.draw_rectangle(0,0,320,20,color=(0,0,0),thickness=-1)
            canvas.draw_string(0,0,"已採集"+str(People_list[FaceNum])+"人臉數據",scale=1,color=(255,255,255),thickness=1)
            Screen.show(canvas,48,8)
            FaceNum = FaceNum + 1
            time.sleep(1)
        faceScore = round(i["recognizeScore"],2)
        if faceScore > 70:
            color = (0,255,0)
            canvas.draw_string(i["X"],i["Y"],str(i["name"])+":"+str(faceScore),scale=1,color=color,thickness=1)
        else:
            color = (255,0,0)
    if key_D.is_pressed() and FaceNum > 0:
        while not key_D.is_pressed()== False:
            time.sleep(0.01)
        FaceRecognition.remove(People_list[len(FaceRecognition)-1]) # 刪除人臉
        canvas.draw_rectangle(0,0,320,20,color=(0,0,0),thickness=-1)
        canvas.draw_string(0,0,"已刪除"+str(People_list[FaceNum-1])+"人臉數據",scale=1,color=(255,255,255),thickness=1)
        Screen.show(canvas,48,8)
        FaceNum = FaceNum - 1
        time.sleep(1)
    Screen.show(canvas,48,8)

數位識別

from AI import Digitalecognition

模型初始化

Digitalecognition()

載入識別

loadModel(img)

參數說明

參數	值	描述
img	圖片	要識別的Image對象，必填

樣例

from cocopiPython import cameraPython,Screen
from AI import Digitalecognition

Screen = Screen() # 畫布初始化
camera = cameraPython() # 攝像頭初始化
Digit = Digitalecognition() # 人臉檢測初始化

while True:
    canvas = camera.capture() # 獲取攝像頭圖像
    canvas = canvas.crop(48,8,224,224) # 裁剪圖像尺寸
    data = Digit.load(canvas)
    for i in data:
        canvas.draw_rectangle(i["X"],i["Y"],i["X"]+i["W"],i["Y"]+i["H"],color=(255,0,0),thickness=1)
        canvas.draw_string(i["X"],i["Y"]-20,(str(i["name"])+":"+str(i["score"])),color=(255,0,0))
    Screen.show(canvas,48,8)

物體檢測

from AI import ObjectDetection

模型初始化

ObjectDetection()

載入識別

loadModel(img)

參數說明

參數	值	描述
img	圖片	要識別的Image對象，必填

樣例

from cocopiPython import cameraPython,Screen
from AI import ObjectDetection

Screen = Screen() # 畫布初始化
camera = cameraPython() # 攝像頭初始化
object = ObjectDetection() # 物體檢測初始化

while True:
    canvas = camera.capture() # 獲取攝像頭圖像
    canvas = canvas.crop(48,8,224,224) # 裁剪圖像尺寸
    data = object.load(canvas)
    for i in data:
        canvas.draw_rectangle(i["X"],i["Y"],i["X"]+i["W"],i["Y"]+i["H"],color=(255,0,0),thickness=1)
        canvas.draw_string(i["X"],i["Y"]-20,(str(i["name"])+":"+str(i["score"])),color=(255,0,0))
    Screen.show(canvas,48,8)

猜拳識別

from AI import RockPaperScissors

模型初始化

RockPaperScissors()

載入識別

loadModel(img)

參數說明

參數	值	描述
img	圖片	要識別的Image對象，必填

樣例

from cocopiPython import cameraPython,Screen
from AI import RockPaperScissors

Screen = Screen()
camera = cameraPython()
RockPaperScissors = RockPaperScissors()

while True:
    canvas = camera.capture()
    canvas = canvas.crop(48,8,224,224)
    data = RockPaperScissors.loadModel(canvas)
    for i in data:
        canvas.draw_rectangle(i["X"],i["Y"],i["X"]+i["W"],i["Y"]+i["H"],color=(255,0,0),thickness=1)
        canvas.draw_string(i["X"],i["Y"]-10,(str(i["name"])+":"+str(i["score"])),color=(255,0,0))
    Screen.show(canvas,48,8)

自學習識別

from AI import SelfLearning

模型初始化

SelfLearning(lables)

參數說明

參數	值	描述
lables	清單	要識別的物體名稱清單，必填

添加类别

addClass(img)

參數說明

參數	值	描述
img	圖片	要識別的Image對象，必填

訓練

trainSave(path)

參數說明

參數	值	描述
path	字串	保存路徑，選填，預設為“/root/user/model/module.bin”

載入識別

loadModel(img,path)

參數說明

參數	值	描述
img	圖片	要識別的Image對象，必填
path	字串	保存路徑，選填，預設為“/root/user/model/module.bin”

樣例

採集

import time
from cocopiPython import cameraPython,Screen
from AI import SelfLearning
from CocoPi import BUTTON

key_C = BUTTON(13)
key_D = BUTTON(7)

Screen = Screen()
camera = cameraPython()
lables=(['滑鼠','鍵盤','水杯'])
SL = SelfLearning(lables)

is_train = 0
count = 0
str_display = "按C鍵添加類別，按D鍵進行學習"

while True:
    # 獲取攝像頭資料
    canvas = camera.capture()
    canvas = canvas.crop(48,8,224,224)
    if key_C.is_pressed():
        while not key_C.is_pressed()== False:
            time.sleep(0.01)
        if count < len(lables):
            # 採集資料添加到清單中
            time.sleep(1)
            SL.addClass(canvas)
            count += 1
    if key_D.is_pressed():
        while not key_D.is_pressed()== False:
            time.sleep(0.01)
        if count >= len(lables) and is_train == 0:
            time.sleep(1)
            # 訓練並保存，參數：path 為保存路徑預設/root/user/model/module.bin
            SL.train()
            is_train = 1
            str_display = "學習已完成"
    canvas.draw_string(0,0,str_display,scale=1,color=(255,0,0),thickness=1)
    canvas.draw_string(0,16,"採集次數："+str(count),scale=1,color=(255,0,0),thickness=1)
    Screen.show(canvas,48,8)

識別

from cocopiPython import cameraPython,Screen
from AI import SelfLearning

Screen = Screen()
camera = cameraPython()
SL= SelfLearning(['滑鼠','鍵盤','水杯'])
SL.load()

while True:
    # 獲取攝像頭資料
    img = camera.capture()
    canvas = img.crop(48,8,224,224)
    data = SL.predict(canvas)
    canvas.draw_string(0,5,(str(data["name"])+":"+str(data["score"])),scale=1,color=(255,0,0),thickness=1)
    Screen.show(canvas,48,8)

拼音識別

from AI import PinyinRecognition

模型初始化

PinyinRecognition(keywords)

參數說明

參數	值	描述
keywords	清單	拼音參數，文字的拼音與音調，必填

獲取結果

getresult()

樣例

from cocopiPython import Screen
from AI import PinyinRecognition

Screen = Screen()
canvas = Screen.create()

keywords = [["ni3 hao3",0.1],["hao3 de1",0.1],["wan3 shang4 hao3",0.1]] # 設置拼音參數
PinyinRecognition = PinyinRecognition(keywords)

data =""
while True:
    canvas.clear()
    data = str(PinyinRecognition.getResult()) # 獲取拼音識別結果
    print("jieguo:"+str(data))
    canvas.draw_string(0,0, "請對準麥克風說出指令", scale = 2, color = (255,0,0) , thickness = 1)
    if data != "None":
        canvas.draw_string(0, 30, "拼音識別結果:", scale=2, color=(255,0,0), thickness = 1)
        canvas.draw_string(30, 60, (data), scale=2, color=(255,255,255), thickness = 1)
    Screen.show(canvas)

車牌識別

from AI import PlateRecognition

模型初始化

PlateRecognition()

載入識別

loadModel(img)

參數說明

參數	值	描述
img	圖片	要識別的Image對象，必填

樣例

from cocopiPython import cameraPython,Screen
from AI import PlateRecognition

Screen = Screen() # 畫布初始化
camera = cameraPython() # 攝像頭初始化
PlateRecognition = PlateRecognition() # 車牌識別初始化
while True:
    canvas = camera.capture() # 獲取攝像頭圖像
    canvas = canvas.crop(48,8,224,224) # 裁剪圖像尺寸
    data = PlateRecognition.load(canvas) # 載入識別
    for i in data:
        canvas.draw_rectangle(i["X"],i["Y"],i["X"]+i["W"],i["Y"]+i["H"],color=(255,0,0),thickness=1) # 檢測框
        canvas.draw_string(i["X"],i["Y"]-10,(str(i["name"])),color=(255,0,0)) # 名稱
    Screen.show(canvas,48,8) # 顯示