chainerのデータセットの作り方 LinearやCNN ref: http://qiita.com/tommyfms2/items/c3fa0cb258c17468cb30

data.txt

...
0,0,11,1,388,484,236,268,500,260,212,392,324,220,216,412,204,244,252,292,4,447,403,589,471,434,448,450,430,410,4,434,448,450,430,410,410,410,410,410,1
0,30,11,0,308,368,324,264,372,384,276,216,372,248,212,192,260,204,208,192,4,434,448,450,430,410,410,410,410,410,4,560,220,238,217,305,267,231,202,185,5
0,0,30,1,216,264,268,236,248,272,244,216,284,236,232,180,280,236,188,188,4,560,220,238,217,305,267,231,202,185,4,305,267,231,202,185,185,185,185,185,3
0,30,5,0,220,192,188,188,184,196,204,184,208,188,188,168,204,200,192,160,4,305,267,231,202,185,185,185,185,185,4,418,418,418,418,418,418,418,418,418,5
...

file1.txt

train, test = chainer.datasets.get_mnist()
print(train[0])

file2.txt

python train_minst.py
(array([ 0.        ,  0.        ,  0.        ,  0.        ,  0.        ,
         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,
         0.71764708,  0.99215692,  0.99215692,  0.81176478,  0.00784314,

... 

         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,
         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,
         0.        ,  0.        ,  0.        ,  0.        ,  0.        ,
         0.        ,  0.        ,  0.        ,  0.        ], dtype=float32), 5)

file3.txt

from chainer.datasets import tuple_dataset
from PIL import Image
import numpy as np
import glob

file6.txt

pathsAndLabels = []
pathsAndLabels.append(np.asarray(["./imageDirectory0/", 0]))
pathsAndLabels.append(np.asarray(["./imageDirectory1/", 1]))
pathsAndLabels.append(np.asarray(["./imageDirectory2/", 2]))

# データを混ぜて、trainとtestがちゃんとまばらになるように。
allData = []
for pathAndLabel in pathsAndLabels:
    path = pathAndLabel[0]
    label = pathAndLabel[1]
    imagelist = glob.glob(path + "*")
    for imgName in imagelist:
        allData.append([imgName, label])
allData = np.random.permutation(allData)

imageData = []
labelData = []
for pathAndLabel in allData:
    img = Image.open(pathAndLabel[0])
    #3チャンネルの画像をr,g,bそれぞれの画像に分ける
    r,g,b = img.split()
    rImgData = np.asarray(np.float32(r)/255.0)
    gImgData = np.asarray(np.float32(g)/255.0)
    bImgData = np.asarray(np.float32(b)/255.0)
    imgData = np.asarray([rImgData, gImgData, bImgData])
    imageData.append(imgData)
    labelData.append(np.int32(pathAndLabel[1]))

threshold = np.int32(len(imageData)/8*7)
train = tuple_dataset.TupleDataset(imageData[0:threshold], labelData[0:threshold])
test  = tuple_dataset.TupleDataset(imageData[threshold:],  labelData[threshold:])

file7.txt

# ここが、上記したポイント部分
# Image.openで取り入れた画像のデータは
# [ [ [r,g,b], [r,g,b], ... ,[r,g,b] ],
#   [ [r,g,b], [r,g,b], ... ,[r,g,b] ],
#   ...
#   [ [r,g,b], [r,g,b], ... ,[r,g,b] ] ]
# のようにrgbがおなじ位置に入ってる感じなのですが、これをそれぞれ赤の画像、緑の画像、青の画像の様にして行列にしないといけません。
# 以下のようなデータにする
# [ [ [ r, r, r, ... ,r ],      ここから
#     ...
#     [ r, r, r, ... ,r ] ],        ここまでが赤の画像
#
#   [ [ g, g, g, ... ,g ],      ここから
#     ...
#     [ g, g, g, ... ,g ] ],        ここまでが緑の画像
#
#     [ b, b, b, ... ,b ],      ここから
#     ...
#     [ b, b, b, ... ,b ] ] ]       ここまでが青の画像

load_dataset.py

train_data = []
train_label = []
data_raw = open("data.txt")
for line in data_raw:
    train = np.array([np.float32(int(x)/255.0) for x in line.split(",")[0:input_num]])
    label = np.int32(line.split(",")[input_num])
    train_data.append(train)
    train_label.append(label)

threshold = np.int32(len(imageData)/10*9)
train = tuple_dataset.TupleDataset(imageData[0:threshold], labelData[0:threshold])
test  = tuple_dataset.TupleDataset(imageData[threshold:],  labelData[threshold:])

train_mnist.py

    ...
    # Load the MNIST dataset
    train, test = chainer.datasets.get_mnist()
    ...