def batch_show(batch,
nr_show=16,
grid_desc=('4v', '4h'),
resize=(600, 800),
title='batch_show'):
"""
Show a batch of images.
:param batch: The batched data: can be either a ndarray of shape (batch_size, h, w, c) or a list
of images.
:param nr_show: Number of images to be displayed. Default set to be 16.
:param grid_desc: Grid description. See `tartist.image.image_grid` for details.
:param resize: Resize factor, a tuple (min_dim, max_dim).
:param title: The title of the shown window.
"""
batch = batch[:nr_show]
batch = np.array(batch)
if len(batch) < 16:
batch = np.concatenate([
batch,
np.zeros([
16 - len(batch), batch.shape[1], batch.shape[2], batch.shape[3]
],
dtype=batch.dtype)
],
axis=0)
img = image.image_grid(batch, grid_desc)
img = image.resize_minmax(img, *resize, interpolation='NEAREST')
image.imshow(title, img)
def vis_and_action(args, cfg, controller, observation):
# here, the observation is actually a state
if hasattr(cfg, 'observe'):
observation = cfg.observe(observation)
display = image.resize_minmax(observation,
*args.winsize,
interpolation='NEAREST')
controller.update(display)
if args.fps > 0:
action = 0
for i in reversed(range(len(cfg.action_keys))):
if controller.test_key(cfg.action_keys[i]):
action = i
break
if controller.test_key(cfg.quit_action_key):
action = __quit_action__
time.sleep((1.0 / args.fps))
else:
action = 0
key = controller.get_last_key()
for i in reversed(range(len(cfg.action_keys))):
if key == cfg.action_keys[i]:
action = i
break
if key == cfg.quit_action_key:
action = __quit_action__
return action
def demo(feed_dict, result, extra_info):
img = result['output'][0, :, :, 0]
img = np.repeat(img[:, :, np.newaxis], 3, axis=2) * 255
img = img.astype('uint8')
img = image.resize_minmax(img, 256)
image.imshow('demo', img)
def main(args, controller):
if args.output is None:
args.output = '{}.conf.py'.format(args.gamename)
if osp.isfile(args.output):
logger.warn('Output config file already exists: {}.'.format(
args.output))
if not yes_or_no('Do you want to overwrite?', default='no'):
controller.quit()
return
game = automake(args.gamename)
game.restart()
action_space = game.action_space
assert isinstance(action_space, DiscreteActionSpace)
action_names = action_space.action_meanings
action_keys = []
logger.critical('All action names: {}'.format(action_names))
logger.critical('Start recording...')
img = image.resize_minmax(game.current_state,
*args.winsize,
interpolation='NEAREST')
controller.update_title(args.gamename)
controller.update(img)
for i in range(len(action_names)):
name = action_names[i]
logger.info('{}-th action, name={}, waiting...'.format(i, name))
action = controller.get_last_key()
action_keys.append(action)
logger.info('{}-th action, name={}, key={}, done.'.format(
i, name, repr(action)))
logger.critical('Recording end.')
logger.info('Recoding quit action key')
quit_key = controller.get_last_key()
logger.info('quit action, key={}, done.'.format(repr(quit_key)))
controller.quit()
with open(args.output, 'w') as f:
f.write('# -*- coding:utf8 -*-\n\n')
f.write(__automake_format__.format(args.gamename))
f.write("name = '{}'\n".format(args.gamename))
f.write("action_names = {}\n".format(repr(action_names)))
f.write("action_keys = {}\n".format(repr(action_keys)))
f.write("quit_action_key = {}\n".format(repr(quit_key)))
f.write("\n")
f.write("# default min size is a tuple (min_dim, max_dim)\n")
f.write("default_winsize = (600, 800)\n")
f.write("# ignore no-action (action=0) in logging\n")
f.write("mute_noaction = True\n")
f.write("\n")
logger.critical('Cfg file written to {}'.format(args.output))
def imshow(img, resize=(600, 800), title='imshow'):
"""
Image show with different parameter order.
:param img: Image.
:param resize: Resize factor, a tuple (min_dim, max_dim).
:param title: The title of the shown window.
"""
img = image.resize_minmax(img, *resize, interpolation='NEAREST')
image.imshow(title, img)
def demo_vae(feed_dict, result, extra_info):
reconstruct = get_env('demo.is_reconstruct', False)
if reconstruct:
img = feed_dict['img'][0, :, :, 0]
omg = result['output'][0, :, :, 0]
img = np.hstack((img, omg))
else:
img = result['output'][0, :, :, 0]
img = np.repeat(img[:, :, np.newaxis], 3, axis=2) * 255
img = img.astype('uint8')
img = image.resize_minmax(img, 256)
image.imshow('demo', img)
def main():
dbdir = osp.realpath(args.output_dir)
io.mkdir(dbdir)
db = LMDBKVStore(dbdir, readonly=False)
files = glob.glob(osp.join(args.input_dir, '*.jpg'))
with db.transaction():
for f in tqdm(files):
img = image.imread(f)
fid = osp.basename(f)[:-4]
img = image.resize_minmax(img, 256, 10000)
img = image.center_crop(img, 256)
db.put(fid, image.jpeg_encode(img))
def demo(feed_dict, result, extra_info):
img = feed_dict['img'][0, :, :, 0]
label = extra_info['label']
img = np.repeat(img[:, :, np.newaxis], 3, axis=2) * 255
img = img.astype('uint8')
img = image.resize_minmax(img, 256)
outputs = [img, np.zeros(shape=[50, 256, 3], dtype='uint8')]
outputs = np.vstack(outputs)
text = 'Pred: {}'.format(result['pred'][0])
text += ' Gt: {}'.format(int(label))
# cv2.putText(outputs, text, (20, 256 + 25), cv2.FONT_HERSHEY_PLAIN, 1.5, (255, 255, 255))
print(text)
image.imshow('demo', outputs)
def demo(feed_dict, result, extra_info):
nr_classes = get_env('dataset.nr_classes')
img = feed_dict['img'][0]
label = extra_info['label']
labels_name = _cifar_labels_name[nr_classes]
img = img.astype('uint8')
img = image.resize_minmax(img, 256)
outputs = [img, np.zeros(shape=[50, 256, 3], dtype='uint8')]
outputs = np.vstack(outputs)
text = 'Pred: {}'.format(labels_name[result['pred'][0]])
text += ' Gt: {}'.format(labels_name[int(label)])
# cv2.putText(outputs, text, (20, 256 + 25), cv2.FONT_HERSHEY_PLAIN, 1.5, (255, 255, 255))
print(text)
image.imshow('demo', outputs)
def demo_draw(feed_dict, result, extra_info):
reconstruct = get_env('demo.is_reconstruct', False)
grid_desc = get_env('demo.draw.grid_desc')
all_outputs = []
for i in range(1000):
name = 'canvas_step{}'.format(i)
if name in result:
all_outputs.append(result[name][0, :, :, 0])
final = image.image_grid(all_outputs, grid_desc)
final = (final * 255).astype('uint8')
if reconstruct:
img = feed_dict['img'][0, :, :, 0]
h = final.shape[0]
w = int(img.shape[1] * h / img.shape[0])
img = (img * 255).astype('uint8')
img = image.resize(img, (h, w))
final = np.hstack((img, final))
final = image.resize_minmax(final, 480, 720)
image.imshow('demo', final)
def r(m):
return np.pad(image.resize_minmax(m.current_state,
128,
interpolation='NEAREST'),
pad_width=[(10, 10), (10, 10), (0, 0)],
mode='constant')