src/separate_pages_ssd/train.py

# -*- coding:utf-8 -*-
import cv2
import os
import keras
from keras.applications.imagenet_utils import preprocess_input
from keras.backend.tensorflow_backend import set_session
from keras.models import Model
from keras.preprocessing import image
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import numpy as np
import pickle
from random import shuffle
from scipy.misc import imread
from scipy.misc import imresize
import tensorflow as tf
import random
from ssd_tools.ssd import SSD300
from ssd_tools.ssd_training import MultiboxLoss
from ssd_tools.ssd_utils import BBoxUtility

#plt.rcParams['figure.figsize'] = (8, 8)
#plt.rcParams['image.interpolation'] = 'nearest'

IMAGE_DIR=os.path.join('training','img')
ANOTATION_FILE=os.path.join('training','page_layout.pkl')

np.set_printoptions(suppress=True)
random.seed(77)
NUM_CLASSES = 2 #4
input_shape = (300, 300, 3)

priors = pickle.load(open(os.path.join('ssd_tools','prior_boxes_ssd300.pkl'), 'rb'))
bbox_util = BBoxUtility(NUM_CLASSES, priors)

gt = pickle.load(open(ANOTATION_FILE, 'rb'))
keys = sorted(gt.keys())
random.shuffle(keys)
num_train = int(round(0.9 * len(keys)))
train_keys = keys[:num_train]
val_keys = keys[num_train:]
num_val = len(val_keys)

class Generator(object):
    def __init__(self, gt, bbox_util,
                 batch_size, path_prefix,
                 train_keys, val_keys, image_size,
                 saturation_var=0.5,
                 brightness_var=0.5,
                 contrast_var=0.5,
                 lighting_std=0.5,
                 hflip_prob=0.5,
                 vflip_prob=0.5,
                 do_crop=True,
                 crop_area_range=[0.8, 1.0],
                 aspect_ratio_range=[1.0,1.0]):
        self.gt = gt
        self.bbox_util = bbox_util
        self.batch_size = batch_size
        self.path_prefix = path_prefix
        self.train_keys = train_keys
        self.val_keys = val_keys
        self.train_batches = len(train_keys)
        self.val_batches = len(val_keys)
        self.image_size = image_size
        self.color_jitter = []
        if saturation_var:
            self.saturation_var = saturation_var
            self.color_jitter.append(self.saturation)
        if brightness_var:
            self.brightness_var = brightness_var
            self.color_jitter.append(self.brightness)
        if contrast_var:
            self.contrast_var = contrast_var
            self.color_jitter.append(self.contrast)
        self.lighting_std = lighting_std
        self.hflip_prob = hflip_prob
        self.vflip_prob = vflip_prob
        self.do_crop = do_crop
        self.crop_area_range = crop_area_range
        self.aspect_ratio_range = aspect_ratio_range

    def grayscale(self, rgb):
        return rgb.dot([0.299, 0.587, 0.114])
        #return rgb.dot([0.333, 0.333, 0.333])

    def saturation(self, rgb):
        gs = self.grayscale(rgb)
        alpha = 2 * np.random.random() * self.saturation_var 
        alpha += 1 - self.saturation_var
        rgb = rgb * alpha + (1 - alpha) * gs[:, :, None]
        return np.clip(rgb, 0, 255)

    def brightness(self, rgb):
        alpha = 2 * np.random.random() * self.brightness_var 
        alpha += 1 - self.saturation_var
        rgb = rgb * alpha
        return np.clip(rgb, 0, 255)

    def contrast(self, rgb):
        gs = self.grayscale(rgb).mean() * np.ones_like(rgb)
        alpha = 2 * np.random.random() * self.contrast_var 
        alpha += 1 - self.contrast_var
        rgb = rgb * alpha + (1 - alpha) * gs
        return np.clip(rgb, 0, 255)

    def lighting(self, img):
        cov = np.cov(img.reshape(-1, 3) / 255.0, rowvar=False)
        eigval, eigvec = np.linalg.eigh(cov)
        noise = np.random.randn(3) * self.lighting_std
        noise = eigvec.dot(eigval * noise) * 255
        img += noise
        return np.clip(img, 0, 255)

    def horizontal_flip(self, img, y):
        if np.random.random() < self.hflip_prob:
            img = img[:, ::-1]
            y[:, [0, 2]] = 1 - y[:, [2, 0]]
        return img, y

    def vertical_flip(self, img, y):
        if np.random.random() < self.vflip_prob:
            img = img[::-1]
            y[:, [1, 3]] = 1 - y[:, [3, 1]]
        return img, y

    def random_sized_crop(self, img, targets):
        img_w = img.shape[1]
        img_h = img.shape[0]
        img_area = img_w * img_h
        random_scale = np.random.random()
        random_scale *= (self.crop_area_range[1] -
                         self.crop_area_range[0])
        random_scale += self.crop_area_range[0]
        target_area = random_scale * img_area
        random_ratio = np.random.random()
        random_ratio *= (self.aspect_ratio_range[1] -
                         self.aspect_ratio_range[0])
        random_ratio += self.aspect_ratio_range[0]
        w = np.round(np.sqrt(target_area * random_ratio))
        h = np.round(np.sqrt(target_area / random_ratio))
        if np.random.random() < 0.5:
            w, h = h, w
        w = min(w, img_w)
        w_rel = w / img_w
        w = int(w)
        h = min(h, img_h)
        h_rel = h / img_h
        h = int(h)
        x = np.random.random() * (img_w - w)
        x_rel = x / img_w
        x = int(x)
        y = np.random.random() * (img_h - h)
        y_rel = y / img_h
        y = int(y)
        img = img[y:y+h, x:x+w]
        new_targets = []
        for box in targets:
            cx = 0.5 * (box[0] + box[2])
            cy = 0.5 * (box[1] + box[3])
            if (x_rel < cx < x_rel + w_rel and
                y_rel < cy < y_rel + h_rel):
                xmin = (box[0] - x_rel) / w_rel
                ymin = (box[1] - y_rel) / h_rel
                xmax = (box[2] - x_rel) / w_rel
                ymax = (box[3] - y_rel) / h_rel
                xmin = max(0, xmin)
                ymin = max(0, ymin)
                xmax = min(1, xmax)
                ymax = min(1, ymax)
                box[:4] = [xmin, ymin, xmax, ymax]
                new_targets.append(box)
        new_targets = np.asarray(new_targets).reshape(-1, targets.shape[1])
        return img, new_targets

    def generate(self, train=True):
        while True:
            if train:
                shuffle(self.train_keys)
                keys = self.train_keys
            else:
                shuffle(self.val_keys)
                keys = self.val_keys
            inputs = []
            targets = []
            for key in keys:
                img_path = self.path_prefix + key
                img = imread(img_path,mode="RGB").astype('float32')
                y = self.gt[key].copy()
                if train and self.do_crop:
                    img, y = self.random_sized_crop(img, y)
                img = imresize(img, self.image_size).astype('float32')
                # boxの位置は正規化されているから画像をリサイズしても
                # 教師信号としては問題ない
                if train:
                    shuffle(self.color_jitter)
                    for jitter in self.color_jitter:
                        img = jitter(img)
                    if self.lighting_std:
                        img = self.lighting(img)
                    if self.hflip_prob > 0:
                        img, y = self.horizontal_flip(img, y)
                    if self.vflip_prob > 0:
                        img, y = self.vertical_flip(img, y)
                # 訓練データ生成時にbbox_utilを使っているのはここだけらしい
                #print(y)
                y = self.bbox_util.assign_boxes(y)
                #print(y)
                inputs.append(img)
                targets.append(y)
                if len(targets) == self.batch_size:
                    tmp_inp = np.array(inputs)
                    #print(tmp_inp.shape)
                    tmp_targets = np.array(targets)
                    inputs = []
                    targets = []
                    yield preprocess_input(tmp_inp), tmp_targets

path_prefix = IMAGE_DIR+"/"
gen = Generator(gt, bbox_util, 5, path_prefix,
                train_keys, val_keys,
                (input_shape[0], input_shape[1]), do_crop=True)

model = SSD300(input_shape, num_classes=NUM_CLASSES)

def schedule(epoch, decay=0.9):
    return base_lr * decay**(epoch)

callbacks = [keras.callbacks.ModelCheckpoint('./checkpoints/weights.{epoch:02d}-{val_loss:.2f}.hdf5',
                                             verbose=1,
                                             save_weights_only=True),
             keras.callbacks.LearningRateScheduler(schedule)]

base_lr = 3e-4
optim = keras.optimizers.Adam(lr=base_lr)
model.compile(optimizer=optim,
              loss=MultiboxLoss(NUM_CLASSES, neg_pos_ratio=5.0).compute_loss)

nb_epoch = 100
history = model.fit_generator(gen.generate(True), gen.train_batches,
                              nb_epoch, verbose=1,
                              callbacks=callbacks,
                              validation_data=gen.generate(False),
                              nb_val_samples=gen.val_batches,
                              nb_worker=1)