train_frcnn.py

from __future__ import division
import random
import pprint
import sys
import time

import boto3
import numpy as np
from optparse import OptionParser
import pickle

sys.path.append('.')

from keras import backend as K
from keras.optimizers import Adam
from keras.layers import Input
from keras.models import Model
from keras_frcnn import config, data_generators
from keras_frcnn import losses as losses
import keras_frcnn.roi_helpers as roi_helpers
from keras.utils import generic_utils
from keras_frcnn.simple_parser import get_data
from keras_frcnn import resnet as nn
from utils.text_classification import process_text_analysis


sys.setrecursionlimit(40000)

parser = OptionParser()

parser.add_option("-p", "--path", dest="train_path", help="Path to training data.", default="annotate.txt")
parser.add_option("-o", "--parser", dest="parser", help="Parser to use. One of simple or pascal_voc",
                  default="simple")
parser.add_option("-n", "--num_rois", type="int", dest="num_rois", help="Number of RoIs to process at once.",
                  default=32)
parser.add_option("--network", dest="network", help="Base network to use. Supports vgg or resnet50.",
                  default='resnet50')
parser.add_option("--hf", dest="horizontal_flips", help="Augment with horizontal flips in training. (Default=false).",
                  action="store_true", default=False)
parser.add_option("--vf", dest="vertical_flips", help="Augment with vertical flips in training. (Default=false).",
                  action="store_true", default=False)
parser.add_option("--rot", "--rot_90", dest="rot_90",
                  help="Augment with 90 degree rotations in training. (Default=false).",
                  action="store_true", default=False)
parser.add_option("--num_epochs", type="int", dest="num_epochs", help="Number of epochs.", default=50)
parser.add_option("--config_filename", dest="config_filename", help=
"Location to store all the metadata related to the training (to be used when testing).",
                  default="config.pickle")
parser.add_option("--output_weight_path", dest="output_weight_path", help="Output path for weights.",
                  default='./model_frcnn.hdf5')
parser.add_option("--input_weight_path", dest="input_weight_path",
                  help="Input path for weights. If not specified, will try to load default weights provided by keras.",
                  default='./model_frcnn.hdf5')

(options, args) = parser.parse_args()

if not options.train_path:  # if filename is not given
    parser.error('Error: path to training data must be specified. Pass --path to command line')

# pass the settings from the command line, and persist them in the config object
C = config.Config()

C.use_horizontal_flips = bool(options.horizontal_flips)
C.use_vertical_flips = bool(options.vertical_flips)
C.rot_90 = bool(options.rot_90)

C.model_path = options.output_weight_path
C.num_rois = int(options.num_rois)
C.network = 'resnet50'

# check if weight path was passed via command line
if options.input_weight_path:
    print('opened input weight')
    C.base_net_weights = options.input_weight_path
else:
    # set the path to weights based on backend and model
    C.base_net_weights = nn.get_weight_path()
    print('new input weight')

all_imgs, classes_count, class_mapping = get_data(options.train_path)

if 'bg' not in classes_count:
    classes_count['bg'] = 0
    class_mapping['bg'] = len(class_mapping)

C.class_mapping = class_mapping

inv_map = {v: k for k, v in class_mapping.items()}

print('Training images per class:')
pprint.pprint(classes_count)
print('Num classes (including bg) = {}'.format(len(classes_count)))

config_output_filename = options.config_filename

with open(config_output_filename, 'wb') as config_f:
    pickle.dump(C, config_f)
    print('Config has been written to {}, and can be loaded when testing to ensure correct results'.format(
        config_output_filename))

random.shuffle(all_imgs)

num_imgs = len(all_imgs)

train_imgs = [s for s in all_imgs if s['imageset'] == 'trainval']
val_imgs = [s for s in all_imgs if s['imageset'] == 'test']

print('Num train samples {}'.format(len(train_imgs)))
print('Num val samples {}'.format(len(val_imgs)))

data_gen_train = data_generators.get_anchor_gt(train_imgs, classes_count, C, nn.get_img_output_length,
                                               K.image_dim_ordering(), mode='train')
data_gen_val = data_generators.get_anchor_gt(val_imgs, classes_count, C, nn.get_img_output_length,
                                             K.image_dim_ordering(), mode='val')

if K.image_dim_ordering() == 'th':
    input_shape_img = (3, None, None)
else:
    input_shape_img = (None, None, 3)

img_input = Input(shape=input_shape_img)
roi_input = Input(shape=(None, 4))
'''
txt_input = Input(shape=(5,))

embedding_dim = 50
maxlen = 5

x = Embedding(1000, embedding_dim, input_length=maxlen)(txt_input)
x = Conv1D(128, 5, activation="relu")(x)
x = GlobalMaxPooling1D()(x)
x = Dense(10, activation='relu')(x)
x = Dense(1, activation='sigmoid')(x)
model_text = Model(txt_input, x)

tokenizer = Tokenizer(num_words=1100)

model_text.compile(optimizer='adam',
                   loss='binary_crossentropy',
                   metrics=['accuracy'])

'''

# define the base network (resnet here, can be VGG, Inception, etc)

print(img_input)
shared_layers = nn.nn_base(img_input, trainable=True)

# define the RPN, built on the base layers
num_anchors = len(C.anchor_box_scales) * len(C.anchor_box_ratios)
rpn = nn.rpn(shared_layers, num_anchors)

classifier = nn.classifier(shared_layers, roi_input, C.num_rois, nb_classes=len(classes_count), trainable=True)

model_rpn = Model(img_input, rpn[:2])
model_classifier = Model([img_input, roi_input], classifier)

# this is a model that holds both the RPN and the classifier, used to load/save weights for the models
model_all = Model([img_input, roi_input], rpn[:2] + classifier)

try:
    print('loading weights from {}'.format(C.base_net_weights))
    model_rpn.load_weights(C.base_net_weights, by_name=True)
    model_classifier.load_weights(C.base_net_weights, by_name=True)
except:
    print('Could not load pretrained model weights. Weights can be found in the keras application folder \
		https://github.com/fchollet/keras/tree/master/keras/applications')
optimizer = Adam(lr=1e-5)
optimizer_classifier = Adam(lr=1e-5)
model_rpn.compile(optimizer=optimizer, loss=[losses.rpn_loss_cls(num_anchors), losses.rpn_loss_regr(num_anchors)])
model_classifier.compile(optimizer=optimizer_classifier,
                         loss=[losses.class_loss_cls, losses.class_loss_regr(len(classes_count) - 1)],
                         metrics={'dense_class_{}'.format(len(classes_count)): 'accuracy'})

model_all.compile(optimizer='sgd', loss='mae')

epoch_length = 1000
num_epochs = int(options.num_epochs)
iter_num = 0

losses = np.zeros((epoch_length, 5))
rpn_accuracy_rpn_monitor = []
rpn_accuracy_for_epoch = []
start_time = time.time()

best_loss = np.Inf

class_mapping_inv = {v: k for k, v in class_mapping.items()}

s3 = boto3.resource('s3')
b_name = 'buttunscreenshot'

print('Starting training')

vis = True


def label_text_buttons(img_data):
    filepath = img_data['filepath']
    img_name = filepath.split('/')[-1]
    res = process_text_analysis(b_name, img_name)
    # img1 = cv2.imread(filepath)

    for b in img_data['bboxes']:
        rx1, ry1, rx2, ry2 = b['x1'], b['y1'], b['x2'], b['y2']
        # cv2.rectangle(img1, (rx1, ry1), (rx2, ry2), (0, 0, 255), 2)
        bt_word = ''
        for word in res:
            dx1, dy1, dx2, dy2 = word[1]
            if dx1 > rx1 and dy1 > ry1 and dx2 < rx2 and dy1 < ry2:
                # cv2.putText(img1, word[0], (rx1, ry1), cv2.FONT_HERSHEY_DUPLEX, 0.5, (0, 0, 255), 1)
                bt_word += f'{word[0]} '
            elif word[0] != ' ' and word[0] not in button_txt and word[0] not in backgraund_txt:
                backgraund_txt.append(word[0])
                # print(f'{word[0]}, bg')

        bt_word = bt_word[:-1]
        if bt_word != ' ' and bt_word not in button_txt:
            button_txt.append(bt_word)
            # print(f'{bt_word}, bt')

    X_text = []
    y_text = []

    for bt in button_txt:
        if bt in backgraund_txt:
            backgraund_txt.remove(bt)
        X_text.append(bt)
        y_text.append(1)

    for bg in backgraund_txt:
        if bg:
            X_text.append(bg)
            y_text.append(0)

    return backgraund_txt, button_txt

'''
text_annotate = False
if text_annotate:
    # annotate text-data
    button_txt = []
    backgraund_txt = []

    for _ in range(len(train_imgs)):
        X, Y, img_data = next(data_gen_train)
        label_text_buttons(img_data)

    for _ in range(len(val_imgs)):
        X, Y, img_data = next(data_gen_val)
        label_text_buttons(img_data)

    text_data = open('/homes/zahara/PycharmProjects/text_button_detection/data/text_bt.txt', 'w')

    for bt in button_txt:
        if bt in backgraund_txt:
            backgraund_txt.remove(bt)
        # print(f'{bt}, bt')
        text_data.write(f'{bt}, bt\n')

    for bg in backgraund_txt:
        if bg:
            text_data.write(f'{bg}, bg\n')
            # print(f'{bg}, bg')

    text_data.close()
    print('Done text annotation')
'''


button_txt = []
backgraund_txt = []

for epoch_num in range(num_epochs):

    progbar = generic_utils.Progbar(epoch_length)
    print('Epoch {}/{}'.format(epoch_num + 1, num_epochs))

    while True:
        try:

            if len(rpn_accuracy_rpn_monitor) == epoch_length and C.verbose:
                mean_overlapping_bboxes = float(sum(rpn_accuracy_rpn_monitor)) / len(rpn_accuracy_rpn_monitor)
                rpn_accuracy_rpn_monitor = []
                print('Average number of overlapping bounding boxes from RPN = {} for {} previous iterations'.format(
                    mean_overlapping_bboxes, epoch_length))
                if mean_overlapping_bboxes == 0:
                    print(
                        'RPN is not producing bounding boxes that overlap the ground truth boxes. Check RPN settings or keep training.')

            X, Y, img_data = next(data_gen_train)

            '''
            X_text, y_text = label_text_buttons(img_data)

            tokenizer.fit_on_texts(X_text)
            X_text = tokenizer.texts_to_sequences(X_text)
            X_text = pad_sequences(X_text, padding='post', maxlen=maxlen)

            loss_txt = model_text.train_on_batch(X_text, y_text)
            pred_text = model_text.predict_on_batch(X_text)
            '''

            loss_rpn = model_rpn.train_on_batch(X, Y)
            P_rpn = model_rpn.predict_on_batch(X)

            R = roi_helpers.rpn_to_roi(P_rpn[0], P_rpn[1], C, K.image_dim_ordering(), use_regr=True, overlap_thresh=0.7,
                                       max_boxes=300)
            # note: calc_iou converts from (x1,y1,x2,y2) to (x,y,w,h) format
            X2, Y1, Y2, IouS = roi_helpers.calc_iou(R, img_data, C, class_mapping)

            if X2 is None:
                rpn_accuracy_rpn_monitor.append(0)
                rpn_accuracy_for_epoch.append(0)
                continue

            neg_samples = np.where(Y1[0, :, -1] == 1)
            pos_samples = np.where(Y1[0, :, -1] == 0)

            if len(neg_samples) > 0:
                neg_samples = neg_samples[0]
            else:
                neg_samples = []

            if len(pos_samples) > 0:
                pos_samples = pos_samples[0]
            else:
                pos_samples = []

            rpn_accuracy_rpn_monitor.append(len(pos_samples))
            rpn_accuracy_for_epoch.append((len(pos_samples)))

            if C.num_rois > 1:
                if len(pos_samples) < C.num_rois // 2:
                    selected_pos_samples = pos_samples.tolist()
                else:
                    selected_pos_samples = np.random.choice(pos_samples, C.num_rois // 2, replace=False).tolist()
                try:
                    selected_neg_samples = np.random.choice(neg_samples, C.num_rois - len(selected_pos_samples),
                                                            replace=False).tolist()
                except ValueError:
                    try:
                        selected_neg_samples = np.random.choice(
                            neg_samples, C.num_rois - len(selected_pos_samples), replace=True).tolist()
                    except:
                        # The neg_samples is [[1 0 ]] only, therefore there's no negative sample
                        continue

                sel_samples = selected_pos_samples + selected_neg_samples
            else:
                # in the extreme case where num_rois = 1, we pick a random pos or neg sample
                selected_pos_samples = pos_samples.tolist()
                selected_neg_samples = neg_samples.tolist()
                if np.random.randint(0, 2):
                    sel_samples = random.choice(neg_samples)
                else:
                    sel_samples = random.choice(pos_samples)

            loss_class = model_classifier.train_on_batch([X, X2[:, sel_samples, :]],
                                                         [Y1[:, sel_samples, :], Y2[:, sel_samples, :]])

            losses[iter_num, 0] = loss_rpn[1]
            losses[iter_num, 1] = loss_rpn[2]

            losses[iter_num, 2] = loss_class[1]
            losses[iter_num, 3] = loss_class[2]
            losses[iter_num, 4] = loss_class[3]

            # losses[iter_num, 5] = loss_txt[1]

            progbar.update(iter_num + 1, [('rpn_cls', losses[iter_num, 0]), ('rpn_regr', losses[iter_num, 1]),
                                          ('detect_cls', losses[iter_num, 2]),
                                          ('detect_regr', losses[iter_num, 3])
                                          # ,('text_regr', losses[iter_num, 5])
                                          ])

            iter_num += 1

            if iter_num == epoch_length:
                loss_rpn_cls = np.mean(losses[:, 0])
                loss_rpn_regr = np.mean(losses[:, 1])
                loss_class_cls = np.mean(losses[:, 2])
                loss_class_regr = np.mean(losses[:, 3])
                class_acc = np.mean(losses[:, 4])
                # loss_txt_cls = np.mean(losses[:, 5])

                mean_overlapping_bboxes = float(sum(rpn_accuracy_for_epoch)) / len(rpn_accuracy_for_epoch)
                rpn_accuracy_for_epoch = []

                if C.verbose:
                    print('Mean number of bounding boxes RPN overlapping ground truth boxes: {}'.format(
                        mean_overlapping_bboxes))
                    print('Accuracy for bounding boxes RPN: {}'.format(class_acc))
                    print('Loss RPN class: {}'.format(loss_rpn_cls))
                    print('Loss RPN regress: {}'.format(loss_rpn_regr))
                    print('Loss Detect class: {}'.format(loss_class_cls))
                    print('Loss Detect regress: {}'.format(loss_class_regr))
                    # print('Loss txt: {}'.format(loss_txt_cls))
                    print('Elapsed time: {} minutes'.format((time.time() - start_time) / 60))

                curr_loss = loss_rpn_cls + loss_rpn_regr + loss_class_cls + loss_class_regr #+ loss_txt_cls
                iter_num = 0
                start_time = time.time()

                if curr_loss < best_loss:
                    if C.verbose:
                        print('Total loss decreased from {} to {}, saving weights'.format(best_loss, curr_loss))
                    best_loss = curr_loss
                    model_all.save_weights(C.model_path)

                break
            break

        except Exception as e:
            print('Exception at end: {}'.format(e))
            continue

print('Training complete, exiting.')