AIはうさぎとアヒルに見える「双安定図形」を判別できるのか？

2023年5月17日2023年5月21日

「双安定図形」ネタばかりで申し訳ございませんが、知覚がコンセプトの会社なのでお許しください

Pythonのコードを共有できそうなので第3段いきます

・第1弾は基礎

アヒルかうさぎか。双安定図形は我々の知覚システムを狂わす。

・第2段は脳波計とニューラルネットワークによる分類の研究の紹介

脳波と人工ニューラルネットワークによる双安定画像の知覚的解釈の分類

・第3段（今回）は検証です

やっていきます

環境

Windows

Anaconda

検証：AIはうさぎとアヒルに見える「双安定図形」を判別できるのか

まずは、download.pyで必要な画像をFlickerAPIを用いてダウンロードします

keyは取得してください

https://www.flickr.com/

download.py

from wave import Wave_write
from sympy import sec
from flickrapi import FlickrAPI
from urllib.request import urlretrieve
from pprint import pprint
import os, time, sys

key = "<FLICKER_API_KEY>"
secret = "de7acaae83fb82c2"
wait_time = 1

animalname = 'rabbit' # ['duck', 'rabbbit'] sys.argv[1]
savedir = "./" + animalname

flickr = FlickrAPI(key, secret, format='parsed-json')
result = flickr.photos.search(
    text = animalname,
    per_page = 400,
    media = 'photos',
    sort = 'relevance',
    safe_search = 1,
    extras = 'url_q, licence'
)

photos = result['photos']
# pprint(photos)

for i, photo in enumerate(photos['photo']):
    url_q = photo['url_q']
    filepath = savedir + '/' + photo['id'] + '.jpg'
    if os.path.exists(filepath): continue
    urlretrieve(url_q, filepath)
    time.sleep(wait_time)

データを作成します

gen_data.py

from PIL import Image
import os, glob
import numpy as np
from sklearn import model_selection
from sklearn.utils import shuffle

classes = ["duck", "rabbit"]
num_classes = len(classes)
image_size = 50
num_testdata = 100

X = []
y = []

for index, classlabel in enumerate(classes):
    photos_dir = "./" + classlabel
    files = glob.glob(photos_dir + "/*.jpg")
    for i, file in enumerate(files):
        if i >= 200:
            break
        image = Image.open(file)
        image = image.convert("RGB")
        image = image.resize((image_size, image_size))
        data = np.asarray(image)
        X.append(data)
        y.append(index)

X = np.array(X, dtype=np.float32)
y = np.array(y, dtype=np.int32)

X, y = shuffle(X, y, random_state=42)

X_train, X_test, y_train, y_test = model_selection.train_test_split(X, y, test_size=0.2, random_state=42)

X_train = X_train / 255.0
X_test = X_test / 255.0

xy = (X_train, X_test, y_train, y_test)
np.save("./animal_aug.npy", xy)

トレーニングします

cnn.py

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, MaxPooling2D
from tensorflow.keras.layers import Activation, Dropout, Flatten, Dense
import numpy as np


num_classes = 2

def main():
    X_train, X_test, y_train, y_test = np.load("./animal_aug.npy", allow_pickle=True)
    y_train = tf.keras.utils.to_categorical(y_train, num_classes)
    y_test = tf.keras.utils.to_categorical(y_test, num_classes)
    model = model_train(X_train, y_train)
    model_eval(model, X_test, y_test)

def model_train(X, y):
    model = Sequential()
    model.add(Conv2D(32, (3, 3), padding='same', input_shape=X.shape[1:]))
    model.add(Activation('relu'))
    model.add(Conv2D(32, (3, 3)))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Dropout(0.25))

    model.add(Conv2D(64, (3, 3), padding='same'))
    model.add(Activation('relu'))
    model.add(Conv2D(64, (3, 3)))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Dropout(0.25))

    model.add(Flatten())
    model.add(Dense(512))
    model.add(Activation('relu'))
    model.add(Dropout(0.5))
    model.add(Dense(num_classes))
    model.add(Activation('softmax'))

    opt = tf.keras.optimizers.RMSprop(learning_rate=0.0001, rho=0.9)
    model.compile(loss='categorical_crossentropy', optimizer=opt, metrics=['accuracy'])
    model.fit(X, y, batch_size=32, epochs=100)

    model.save('./animal_cnn_aug.h5')

    return model

def model_eval(model, X, y):
    scores = model.evaluate(X, y, verbose=1)
    print('Test Loss:', scores[0])
    print('Test Accuracy:', scores[1])

if __name__ == "__main__":
    main()

予測します

predict.py

import tensorflow as tf
from keras.models import Sequential, load_model
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import Activation, Dropout, Flatten, Dense
from keras.utils import np_utils
import keras, sys
import numpy as np
from PIL import Image

classes = ["duck", "rabbit"]
num_classes = len(classes)
image_size = 50

def build_model():
    model = Sequential()
    model.add(Conv2D(32,(3,3), padding='same',input_shape=(50,50,3)))
    model.add(Activation('relu'))
    model.add(Conv2D(32,(3,3)))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2,2)))
    model.add(Dropout(0.25))

    model.add(Conv2D(64,(3,3), padding='same'))
    model.add(Activation('relu'))
    model.add(Conv2D(64,(3,3)))
    model.add(Activation('relu'))
    model.add(MaxPooling2D(pool_size=(2,2)))
    model.add(Dropout(0.25))

    model.add(Flatten())
    model.add(Dense(512))
    model.add(Activation('relu'))
    model.add(Dropout(0.5))
    model.add(Dense(2))
    model.add(Activation('softmax'))

    opt = tf.keras.optimizers.RMSprop(learning_rate=0.0001)

    model.compile(loss='categorical_crossentropy',optimizer=opt,metrics=['accuracy'])

    model = load_model('./animal_cnn_aug.h5')

    return model

def main():
    image = Image.open(sys.argv[1])
    image = image.convert('RGB')
    image = image.resize((image_size,image_size))
    data = np.asarray(image)
    X = []
    X.append(data)
    X = np.array(X)
    model = build_model()

    result = model.predict([X])[0]
    predicted = result.argmax()
    percentage = int(result[predicted] * 100)
    print("{0} ({1} %)".format(classes[predicted], percentage))

if __name__ == '__main__':
    main()

結果を見ましょう

手順は以下です