def save_focus_img(dataset, all_focus, save_path, changepoints=[]):
focus_img = util.extract_neighbor(dataset,
all_focus,
range(len(all_focus)),
nb_size=(15, 20))
for i, img in enumerate(focus_img):
# focused neighbor image
file_name = 'focus_img_%d.png' % (i)
file_path = os.path.join(save_path, file_name)
imsave(file_path, img)
print('saved under', save_path)
pos_cnt = np.zeros(dataset.get_shape()[-2:])
cp_mask = np.zeros(len(all_focus), dtype=bool)
cp_mask[changepoints] = True
save_subpath = util.get_dir(os.path.join(save_path, 'marker'))
for i, img in enumerate(dataset.get_frame(0, len(all_focus))):
# for i, img in enumerate(util.remove_mean(dataset.get_frame(0, len(all_focus)), all_focus)):
# marker
file_name = 'marker_%d.png' % (i)
marker_file_path = os.path.join(save_subpath, file_name)
marker_pos = (all_focus[i] * dataset.get_shape()[2:]).astype(int)
img[0, marker_pos[0], :] = 1
img[0, :, marker_pos[1]] = 1
if cp_mask[i]:
img = 1 - img
pos_cnt[marker_pos[0], marker_pos[1]] += 1
imsave(marker_file_path, img[0])
print('focus by marker saved under', save_subpath)
if np.sum(pos_cnt) > 0:
pos_cnt = np.sqrt(1.0 - (pos_cnt / np.max(pos_cnt) - 1)**2)
pos_cnt_file_path = os.path.join(save_path, 'counter_pos.png')
imsave(pos_cnt_file_path, pos_cnt)
print('position count saved at', pos_cnt_file_path)
def save_imgs(imgs, save_path):
save_subpath = util.get_dir(os.path.join(save_path, 'intensity'))
for i, img in enumerate(imgs):
file_name = 'intensity_%d.png' % (i)
file_path = os.path.join(save_subpath, file_name)
imsave(file_path, util.feature_normalize(img[0]))
print('focus intensity saved under', save_subpath)
def save_imgs(imgs, save_path, title_extra=False):
save_subpath = util.get_dir(os.path.join(save_path, 'intensity'))
for i, img in enumerate(imgs):
if title_extra:
file_name = 'intensity_%d_%.2f.png'%(i, title_extra[i])
else:
file_name = 'intensity_%d.png'%(i)
file_path = os.path.join(save_subpath, file_name)
util.imsave(file_path, util.feature_normalize(img[0]))
print('focus intensity saved under', save_subpath)
def load_model(prefix, model_id, net_params, *args, **kwargs):
params = load_param((util.get_dir(prefix), model_id))
model = ModelFocusCNN(
image_shape=(84, 84),
net_params=net_params,
*args,
**kwargs,
)
model.set_parameters(params)
return model
def __init__(self, dim, *args, **kwargs):
self.dim = dim
self.id = kwargs.get('id', str(np.random.randint(42000, 42999)))
self.cmaes_params = kwargs.get('cmaes_params', {})
self.save_path = util.get_dir(kwargs.get('save_path',
os.path.join('cmaes_soln', self.id)))
self.max_niter = kwargs.get('max_niter', 100)
self.nproc = kwargs.get('nproc', 1)
self.cheating = kwargs.get('cheating', None)
self.file_id = 0
# initialize CMA-ES core
self.reset()
logger.info('created wrapper with id= %s', self.id)
def __init__(self, game_name, Actor, n_state, save_path, *args, **kwargs):
# create OpenAI Atari
self.save_path = util.get_dir(save_path)
self.atari_env = SubprocVecEnv(
[make_env('BreakoutNoFrameskip-v4', 1234, 0, self.save_path)])
self.frame_shape = self.atari_env.observation_space.shape[-2:]
self.n_state = n_state
self.binarize = kwargs.get('binarize', None)
# actor for auto run
self.action_space = self.atari_env.action_space.n
self.actor = Actor(self.action_space)
# TODO: online generatation
self._generate_all_states()
else:
print('using simple linear changepoint detector')
cpd = LinearCPD(np.pi/4.0)
"""
Load model
"""
net_params = json.loads(open(args.net).read())
r_model = load_model(
prefix,
model_id,
net_params=net_params,
use_prior=args.prior,
argmax_mode=args.argmax_mode,
)
save_path = util.get_dir(os.path.join(prefix, 'focus_img_%s'%model_id))
if plot_flags['plot_filter']:
plot_model_filter(r_model, save_path)
# boosting with trained models
if args.boost:
# partial ball model to be boosted
ball_model_id = 'results/cmaes_soln/focus_atari_breakout/42080_16.npy'
ball_net_params_text = open('objRecog/net_params/two_layer.json').read()
ball_net_params = json.loads(ball_net_params_text)
ball_params = load_param(ball_model_id)
ball_model = ModelFocusCNN(
image_shape=(84, 84),
net_params=ball_net_params,
)
ball_model.set_parameters(ball_params)
