File size: 10,942 Bytes
6e5cc8b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
# coding=utf-8
# Copyright 2021 The Google Research Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

# Lint as: python3
"""A wrapper for dm_control environments which applies color distractions."""
import os

from PIL import Image
import collections
from dm_control.rl import control
import numpy as np

import tensorflow as tf
from dm_control.mujoco.wrapper import mjbindings

DAVIS17_TRAINING_VIDEOS = [
    'bear', 'bmx-bumps', 'boat', 'boxing-fisheye', 'breakdance-flare', 'bus',
    'car-turn', 'cat-girl', 'classic-car', 'color-run', 'crossing',
    'dance-jump', 'dancing', 'disc-jockey', 'dog-agility', 'dog-gooses',
    'dogs-scale', 'drift-turn', 'drone', 'elephant', 'flamingo', 'hike',
    'hockey', 'horsejump-low', 'kid-football', 'kite-walk', 'koala',
    'lady-running', 'lindy-hop', 'longboard', 'lucia', 'mallard-fly',
    'mallard-water', 'miami-surf', 'motocross-bumps', 'motorbike', 'night-race',
    'paragliding', 'planes-water', 'rallye', 'rhino', 'rollerblade',
    'schoolgirls', 'scooter-board', 'scooter-gray', 'sheep', 'skate-park',
    'snowboard', 'soccerball', 'stroller', 'stunt', 'surf', 'swing', 'tennis',
    'tractor-sand', 'train', 'tuk-tuk', 'upside-down', 'varanus-cage', 'walking'
]
DAVIS17_VALIDATION_VIDEOS = [
    'bike-packing', 'blackswan', 'bmx-trees', 'breakdance', 'camel',
    'car-roundabout', 'car-shadow', 'cows', 'dance-twirl', 'dog', 'dogs-jump',
    'drift-chicane', 'drift-straight', 'goat', 'gold-fish', 'horsejump-high',
    'india', 'judo', 'kite-surf', 'lab-coat', 'libby', 'loading', 'mbike-trick',
    'motocross-jump', 'paragliding-launch', 'parkour', 'pigs', 'scooter-black',
    'shooting', 'soapbox'
]
SKY_TEXTURE_INDEX = 0
Texture = collections.namedtuple('Texture', ('size', 'address', 'textures'))


def imread(filename):
    img = Image.open(filename)
    img_np = np.asarray(img)
    return img_np


def size_and_flatten(image, ref_height, ref_width):
    # Resize image if necessary and flatten the result.
    image_height, image_width = image.shape[:2]

    if image_height != ref_height or image_width != ref_width:
        image = tf.cast(tf.image.resize(image, [ref_height, ref_width]), tf.uint8)
    return tf.reshape(image, [-1]).numpy()


def blend_to_background(alpha, image, background):
    if alpha == 1.0:
        return image
    elif alpha == 0.0:
        return background
    else:
        return (alpha * image.astype(np.float32)
                + (1. - alpha) * background.astype(np.float32)).astype(np.uint8)


class DistractingBackgroundEnv(control.Environment):
    """Environment wrapper for background visual distraction.

  **NOTE**: This wrapper should be applied BEFORE the pixel wrapper to make sure
  the background image changes are applied before rendering occurs.
  """

    def __init__(self,
                 env,
                 dataset_path=None,
                 dataset_videos=None,
                 video_alpha=1.0,
                 ground_plane_alpha=1.0,
                 num_videos=None,
                 dynamic=False,
                 seed=None,
                 shuffle_buffer_size=None,
                 fixed=False,
                 ):

        if not 0 <= video_alpha <= 1:
            raise ValueError('`video_alpha` must be in the range [0, 1]')

        self._env = env
        self._video_alpha = video_alpha
        self._ground_plane_alpha = ground_plane_alpha
        self._random_state = np.random.RandomState(seed=seed)
        self._dynamic = dynamic
        self._shuffle_buffer_size = shuffle_buffer_size
        self._background = None
        self._current_img_index = 0

        if not dataset_path or num_videos == 0:
            # Allow running the wrapper without backgrounds to still set the ground
            # plane alpha value.
            self._video_paths = []
        else:
            # Use all videos if no specific ones were passed.
            if not dataset_videos:
                dataset_videos = sorted(tf.io.gfile.listdir(dataset_path))
            # Replace video placeholders 'train'/'val' with the list of videos.
            elif dataset_videos in ['train', 'training']:
                dataset_videos = DAVIS17_TRAINING_VIDEOS
            elif dataset_videos in ['val', 'validation']:
                dataset_videos = DAVIS17_VALIDATION_VIDEOS
            # Get complete paths for all videos.
            video_paths = [
                os.path.join(dataset_path, subdir) for subdir in dataset_videos
            ]

            # Optionally use only the first num_paths many paths.
            if num_videos is not None:
                if num_videos > len(video_paths) or num_videos < 0:
                    raise ValueError(f'`num_bakground_paths` is {num_videos} but '
                                     'should not be larger than the number of available '
                                     f'background paths ({len(video_paths)}) and at '
                                     'least 0.')
                video_paths = video_paths[:num_videos]

            self._video_paths = video_paths

        self.fixed = fixed
        self.first_reset = False

    def reset(self):
        """Reset the background state."""
        time_step = self._env.reset()
        if not self.fixed or not self.first_reset:
            self._reset_background()
            self.first_reset = True
        return time_step

    def _reset_background(self):
        # Make grid semi-transparent.
        if self._ground_plane_alpha is not None:
            self._env.physics.named.model.mat_rgba['grid',
                                                   'a'] = self._ground_plane_alpha

        # For some reason the height of the skybox is set to 4800 by default,
        # which does not work with new textures.
        self._env.physics.model.tex_height[SKY_TEXTURE_INDEX] = 800

        # Set the sky texture reference.
        sky_height = self._env.physics.model.tex_height[SKY_TEXTURE_INDEX]
        sky_width = self._env.physics.model.tex_width[SKY_TEXTURE_INDEX]
        sky_size = sky_height * sky_width * 3
        sky_address = self._env.physics.model.tex_adr[SKY_TEXTURE_INDEX]

        sky_texture = self._env.physics.model.tex_rgb[sky_address:sky_address +
                                                                  sky_size].astype(np.float32)

        if self._video_paths:
            if self._shuffle_buffer_size:
                # Shuffle images from all videos together to get background frames.
                file_names = [
                    os.path.join(path, fn)
                    for path in self._video_paths
                    for fn in tf.io.gfile.listdir(path)
                ]
                self._random_state.shuffle(file_names)
                # Load only the first n images for performance reasons.
                file_names = file_names[:self._shuffle_buffer_size]
                images = [imread(fn) for fn in file_names]
            else:
                # Randomly pick a video and load all images.
                video_path = self._random_state.choice(self._video_paths)
                file_names = tf.io.gfile.listdir(video_path)
                if not self._dynamic:
                    # Randomly pick a single static frame.
                    # file_names = [self._random_state.choice(file_names)]
                    file_names.sort()
                    file_names = [file_names[0]]
                images = [imread(os.path.join(video_path, fn)) for fn in file_names]

            # Pick a random starting point and steping direction.
            self._current_img_index = self._random_state.choice(len(images))
            self._step_direction = self._random_state.choice([-1, 1])

            # Prepare images in the texture format by resizing and flattening.

            # Generate image textures.
            texturized_images = []
            for image in images:
                image_flattened = size_and_flatten(image, sky_height, sky_width)
                new_texture = blend_to_background(self._video_alpha, image_flattened,
                                                  sky_texture)
                texturized_images.append(new_texture)

        else:
            self._current_img_index = 0
            texturized_images = [sky_texture]

        self._background = Texture(sky_size, sky_address, texturized_images)
        self._apply()

    def step(self, action):
        time_step = self._env.step(action)

        if time_step.first():
            self._reset_background()
            return time_step

        if self._dynamic and self._video_paths:
            # Move forward / backward in the image sequence by updating the index.
            self._current_img_index += self._step_direction

            # Start moving forward if we are past the start of the images.
            if self._current_img_index <= 0:
                self._current_img_index = 0
                self._step_direction = abs(self._step_direction)
            # Start moving backwards if we are past the end of the images.
            if self._current_img_index >= len(self._background.textures):
                self._current_img_index = len(self._background.textures) - 1
                self._step_direction = -abs(self._step_direction)

            self._apply()
        return time_step

    def _apply(self):
        """Apply the background texture to the physics."""

        if self._background:
            start = self._background.address
            end = self._background.address + self._background.size
            texture = self._background.textures[self._current_img_index]

            self._env.physics.model.tex_rgb[start:end] = texture
            # Upload the new texture to the GPU. Note: we need to make sure that the
            # OpenGL context belonging to this Physics instance is the current one.
            with self._env.physics.contexts.gl.make_current() as ctx:
                ctx.call(
                    mjbindings.mjlib.mjr_uploadTexture,
                    self._env.physics.model.ptr,
                    self._env.physics.contexts.mujoco.ptr,
                    SKY_TEXTURE_INDEX,
                )

    # Forward property and method calls to self._env.
    def __getattr__(self, attr):
        if hasattr(self._env, attr):
            return getattr(self._env, attr)
        raise AttributeError("'{}' object has no attribute '{}'".format(
            type(self).__name__, attr))