scikit-learnを使った手書き数字認識

Question

tensorflowをいきなり使うのは難しかったので、scikit-learnを使い数字認識を作ろうと思いました。

手順としては
1.pythonファイルを実行するとペイントソフトが起動する
2.数字を描画する
3.画像として保存
4.数字を学習したAIがその画像を認識する
5.結果を出力

今回の参考サイトは
ペイントソフトを作って、ついでに手書き数字認識もする

基本的にサイトに書かれていることしかしていないので、以下のエラーが一体何なのかわからず困っています。
K近傍法とやらのエラーっぽい感じなのはわかるのですが・・・。

インデント修正、<の変換、色指定（COLOR_WHITE）のエラーはどうにか調べて修正し、現在描画して画像を保存するところまではできています。
今回分からないエラーは

Image has saved correctly.
C:\Users\ユーザー名\AppData\Local\Programs\Python\Python36\lib\site-packages\sklearn\utils\deprecation.py:144: FutureWarning: The sklearn.neighbors.classification module is  deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.neighbors. Anything that cannot be imported from sklearn.neighbors is now part of the private API.
  warnings.warn(message, FutureWarning)
C:\Users\ユーザー名\AppData\Local\Programs\Python\Python36\lib\site-packages\sklearn\utils\deprecation.py:144: FutureWarning: The sklearn.neighbors.kd_tree module is  deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.neighbors. Anything that cannot be imported from sklearn.neighbors is now part of the private API.
  warnings.warn(message, FutureWarning)
C:\Users\ユーザー名\AppData\Local\Programs\Python\Python36\lib\site-packages\sklearn\utils\deprecation.py:144: FutureWarning: The sklearn.neighbors.dist_metrics module is  deprecated in version 0.22 and will be removed in version 0.24. The corresponding classes / functions should instead be imported from sklearn.neighbors. Anything that cannot be imported from sklearn.neighbors is now part of the private API.
  warnings.warn(message, FutureWarning)
C:\Users\ユーザー名\AppData\Local\Programs\Python\Python36\lib\site-packages\sklearn\base.py:318: UserWarning: Trying to unpickle estimator KNeighborsClassifier from version 0.21.3 when using version 0.22.1. This might lead to breaking code or invalid results. Use at your own risk.
  UserWarning)
Traceback (most recent call last):
  File "paint.py", line 87, in <module>
    App()
  File "paint.py", line 31, in __init__
    pyxel.run(self.update, self.draw)
  File "C:\Users\ユーザー名\AppData\Local\Programs\Python\Python36\lib\site-packages\pyxel\app.py", line 255, in run
    main_loop()
  File "C:\Users\ユーザー名\AppData\Local\Programs\Python\Python36\lib\site-packages\pyxel\app.py", line 247, in main_loop
    self._update_frame()
  File "C:\Users\ユーザー名\AppData\Local\Programs\Python\Python36\lib\site-packages\pyxel\app.py", line 438, in _update_frame
    self._update()
  File "paint.py", line 57, in update
    self.pred_digit = model.load_predict()
  File "C:\Users\ユーザー名\Desktop\AI\pyxelDigitRecognition\model.py", line 43, in load_predict
    pred = loaded_model.predict(img)[0] # np.array to int
  File "C:\Users\ユーザー名\AppData\Local\Programs\Python\Python36\lib\site-packages\sklearn\neighbors\_classification.py", line 173, in predict
    neigh_dist, neigh_ind = self.kneighbors(X)
  File "C:\Users\ユーザー名\AppData\Local\Programs\Python\Python36\lib\site-packages\sklearn\neighbors\_base.py", line 580, in kneighbors
    n_samples_fit = self.n_samples_fit_
AttributeError: 'KNeighborsClassifier' object has no attribute 'n_samples_fit_'

コード

# 3rd party
import numpy as np
from PIL import Image
import pyxel

# my library
import model

# constants
WINDOW_SIZE = 64


# ===== main =====
class App:

    # 0: white, -1: black
    windowData = [[0]*WINDOW_SIZE for _ in range(WINDOW_SIZE)]
    pred_digit = None

    def __init__(self):
        pyxel.init(WINDOW_SIZE, WINDOW_SIZE)
        pyxel.mouse(visible=True)
        pyxel.run(self.update, self.draw)


    def update(self):
        if pyxel.btn(pyxel.KEY_DELETE):
            self.pred_digit = None
            # Change each pixel's color to white
            for y in range(pyxel.height):
                for x in range(pyxel.width):
                    self.windowData[y][x] = 0

        if pyxel.btnp(pyxel.MOUSE_LEFT_BUTTON, hold=2, period=1):
            # Draw a line
            # This change pixel's colors like a star(*)
            # (the center is (mouse_x,mouse_y))
            x, y = pyxel.mouse_x, pyxel.mouse_y
            dx = [-3,-2,-1, 0, 0, 0,0,0,0,0,1,2,3,-1,-1,1,1,-2,-2,2,2]
            dy = [ 0, 0, 0,-3,-2,-1,0,1,2,3,0,0,0,-1,1,-1,1,2,-2,2,-2]
            for i in range(len(dx)):
                nx, ny = x+dx[i], y+dy[i]
                if 0 <= nx < pyxel.width and 0 <= ny < pyxel.height:
                    self.windowData[ny][nx] = -1

        if pyxel.btn(pyxel.KEY_S):
            # save image and recognize the digit
            self._saveImage()
            self.pred_digit = model.load_predict()


    def draw(self):
        pyxel.cls(7)
        # show window
        for y in range(pyxel.height):
            for x in range(pyxel.width):
                if self.windowData[y][x]==-1:
                    pyxel.pix(x, y, 0)

        pyxel.text(0, 0, 'PREDICT: {}'.format(self.pred_digit), 8)


    def _saveImage(self):
        """ save 64x64 window's image into 8x8 picture
        The file name is 'screen_shot.png'
        """
        img = Image.new('RGB', (pyxel.width, pyxel.height))
        for y in range(pyxel.height):
            for x in range(pyxel.width):
                if self.windowData[y][x]==0:
                    img.putpixel((x,y), (255,255,255))
                else:
                    img.putpixel((x,y), (0,0,0))
        img = img.resize((8, 8), Image.BICUBIC)
        img.save("images/screen_shot.png")
        print("Image has saved correctly.")


App()

import numpy as np
import pickle
from PIL import Image
from sklearn.datasets import load_digits
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier


def train():
    """Train k-nearest neighbors model with Digits dataset
    """
    digits = load_digits()
    X = digits.data
    y = digits.target

    # print(X.shape) # (1797, 64)
    # print(y.shape) # (1797,)

    X_train,X_test,y_train,y_test = train_test_split(X, y)

    knn = KNeighborsClassifier()
    knn.fit(X_train, y_train)
    # print(knn.score(X_test, y_test)) # 0.98 lol

    # save model
    with open('knn_digit.pkl', 'wb') as f:
        pickle.dump(knn, f)


def load_predict() -> int:
    """load a trained model and predict
    """

    with open('knn_digit.pkl', 'rb') as f:
        loaded_model = pickle.load(f)

    # Open image and extract features
    img = Image.open("images/screen_shot.png")
    img = img.convert('L') # convert (r,g,b) to gray scale (0-255)
    img = (255 - np.array(img))//16 + 1 # convert to 0-15
    img = img.reshape(1, 64)

    pred = loaded_model.predict(img)[0] # np.array to int
    # print("PREDICT: ", pred)
    return pred

ライブラリなど

windows10
python3.6
pyxel
scikit-learn 0.22

Comments

[hayataka2049]

情報が不足しています。完全なエラーメッセージとコードをご提示ください。

Answer 1

scikit-learnのバージョンを0.21.3にすることで成功しました。