【Python】画像認識 - データ読み込み方法まとめ 【DeepLearning】

データセットがカテゴリ（ラベル）毎にフォルダ分けされている場合

BASE_DATA_FOLDER = "../input/"
TRAin_DATA_FOLDER = os.path.join(BASE_DATA_FOLDER, "train")
images_per_class = {}
for class_folder_name in os.listdir(TRAin_DATA_FOLDER):
    class_folder_path = os.path.join(TRAin_DATA_FOLDER, class_folder_name)
    class_label = class_folder_name
    images_per_class[class_label] = []
    for image_path in glob(os.path.join(class_folder_path, "*.png")):
        image_bgr = cv2.imread(image_path, cv2.IMREAD_COLOR)
        images_per_class[class_label].append(image_bgr)

z = glob.glob("../input/train/*/*.png")
ori_label = []
ori_imgs = []
for fn in z:
    if fn[-3:] != 'png':
        continue
    ori_label.append(fn.split('/')[-2])
    new_img = Image.open(fn)
    ori_imgs.append(ImageOps.fit(new_img, (48, 48), Image.ANTIALIAS).convert('RGB'))

#画像のサイズ指定
ScaleTo = 70
seed = 7

#トレーニングデータの読み込み
path = "./train/*/*.png"
files = glob(path)

trainImg = []
trainLabel = []
j = 1
num = len(files)

for img in files:
    print(str(j) + "/" + str(num) , end="\r")
    trainImg.append(cv2.resize(cv2.imread(img) ,(ScaleTo,ScaleTo)))
    trainLabel.append(img.split("/")[-2])
    j += 1

trainImg = np.asarray(trainImg)
trainLabel = pd.DataFrame(trainLabel)

ImageDataBunchの作成

#トレーニングデータの読み込み
data_path = Path('../input')
df = pd.read_csv(data_path/'train.csv')
df.head()


#提出用データの読み込み
sub_csv = pd.read_csv(data_path/'sample_submission.csv')
sub_csv.head()

#データ束の作成
def create_databunch(valid_idx):
    test = ImageList.from_df(sub_csv, path=data_path/'test', folder='test')
    data = (ImageList.from_df(df, path=data_path/'train', folder='train')
            .split_by_idx(valid_idx)
            .label_from_df()
            .add_test(test)
            .transform(get_transforms(flip_vert=True, max_rotate=20.0), size=128)
            .databunch(path='.', bs=64)
            .normalize(imagenet_stats)
           )
    return data

画像データとラベルが別々に保存されている場合

#トレーニングデータ、テストデータ、ラベルデータのパス読み込み
train_dir = "../input/train/train/"
test_dir = "../input/test/test/"
train_df = pd.read_csv('../input/train.csv')
train_df.head()


X_tr = []
Y_tr = []
images = train_df["id"].values
for img_id in tqdm_notebook(images):
    X_tr.append(cv2.imread(train_dir+img_id))
    Y_tr.append(train_df[train_df["id"]==img_id]["has_cactus"].values[0])
X_tr = np.asarray(X_tr)
X_tr = X_tr.astype('float32')
X_tr /= 255
Y_tr = np.asarray(Y_tr)

Zip以下に配置されたフォルダに画像がある場合

#ファイルディレクトリの指定
data_folder = Path("../input")
train_df = pd.read_csv("../input/train.csv")
test_df = pd.read_csv("../input/sample_submission.csv")



#学習用データの読み込み
train_img = (ImageList.from_df(train_df,path=data_folder/"train",folder="train")
            .split_by_rand_pct(0.01)
            .label_from_df()
            .add_test(test_img)
            .transform(trfm,size=128)
            .databunch(path=".",bs=64,device=torch.device('cuda:0'))
             .normalize(imagenet_stats))