chainerの作法 その9

概要

chainerの作法を調べてみた。
手書きを推定してみた。

写真

サンプルコード

import sys
import numpy as np
from PyQt5.QtWidgets import *
from PyQt5.QtGui import *
from PyQt5.QtCore import *
from PIL import Image
import chainer
import chainer.functions as F
import chainer.links as L
from chainer import training, Chain, datasets, serializers
from chainer.training import extensions
import PIL

class MLP(Chain):
	def __init__(self, n_units, n_out):
		super(MLP, self).__init__(l1 = L.Linear(None, n_units), l2 = L.Linear(None, n_units), l3 = L.Linear(None, n_out))
	def __call__(self, x):
		h1 = F.relu(self.l1(x))
		h2 = F.relu(self.l2(h1))
		return self.l3(h2)

class pline(QWidget):
	def __init__(self, parent = None):
		QWidget.__init__(self, parent)
		self.px = None
		self.py = None
		self.points = []
		self.psets = []
		self.points_saved = []
		self.image2 = QImage(180, 180, QImage.Format_RGBA8888)
		self.image2.fill(qRgb(0, 0, 0))
	def mousePressEvent(self, event):
		self.points.append(event.pos())
		self.update()
	def mouseMoveEvent(self, event):
		self.points.append(event.pos())
		self.update()
	def mouseReleaseEvent(self, event):
		self.pressed = False
		self.psets.append(self.points)
		self.points_saved.extend(self.points)
		self.points = []
		self.update()
	def paintEvent(self, event):
		painter = QPainter(self)
		painter.setPen(QPen(Qt.black, 8, Qt.SolidLine))
		for points in self.psets:
			painter.drawPolyline(*points)
		if self.points:
			painter.drawPolyline(*self.points)
		painter2 = QPainter(self.image2)
		painter2.setPen(QPen(Qt.white, 8, Qt.SolidLine))
		for points in self.psets:
			painter2.drawPolyline(*points)
		if self.points:
			painter2.drawPolyline(*self.points)
	def save(self):
		self.image2.save(r'd:/test.png', 'png')
	def clear(self):
		self.points = []
		self.psets = []
		self.points_saved = []
		self.image2 = QImage(180, 180, QImage.Format_RGBA8888)
		self.image2.fill(qRgb(0, 0, 0))
		self.repaint()

class MainWindow(QWidget):
	def __init__(self, parent = None):
		super(MainWindow, self).__init__(parent)
		Button0 = QPushButton("recog")
		Button0.clicked.connect(self.recog)
		self.pain = pline()
		self.setGeometry(200, 200, 200, 200)
		layout = QVBoxLayout()
		layout.addWidget(self.pain)
		layout.addWidget(Button0)
		self.setLayout(layout)
		self.image = np.zeros((paint_width, paint_height, 3), np.uint8)
		self.setWindowTitle('MNIST')
	def recog(self):
		self.pain.save()
		img = Image.open('d:/test.png')
		img = img.resize((28, 28))
		img = img.convert('L')
		img_arr = np.array(img, dtype = np.float32).reshape(1, -1) / 255
		x = img_arr
		y = model.predictor(x)
		y = y.data
		p = y.argmax(axis = 1)[0]
		msg = QMessageBox()
		msg.setText("推定は、" + str(p))
		msg.setWindowTitle("recog")
		msg.setStandardButtons(QMessageBox.Ok)
		retval = msg.exec_()
		self.pain.clear()

paint_width = 180
paint_height = 180
model = L.Classifier(MLP(1000, 10))
serializers.load_npz('mnist3.model', model)

if __name__ == '__main__':
	app = QApplication(sys.argv)
	main_window = MainWindow()
	main_window.show()
	sys.exit(app.exec_())

以上。