def __init__(self, method):
if not isinstance(method, str):
raise ValueError(f'Invalid method type "{type(method)}"')
from compressai import available_entropy_coders
if method not in available_entropy_coders():
methods = ", ".join(available_entropy_coders())
raise ValueError(f'Unknown entropy coder "{method}"'
f" (available: {methods})")
if method == "ans":
from compressai import ans
encoder = ans.RansEncoder()
decoder = ans.RansDecoder()
elif method == "rangecoder":
import range_coder
encoder = range_coder.RangeEncoder()
decoder = range_coder.RangeDecoder()
self.name = method
self._encoder = encoder
self._decoder = decoder
def compress_image(self,
code_list: List[torch.Tensor],
savepath="default.dci"):
"""
Args:
code_list: List of encodings for image
savepath: path to save the image to
Returns: Nothing, writes compressed image savepath
"""
if not self.cdf:
raise Exception(
"Cannot compress image without cdf function (hint: call build_latent_distribution)"
)
flat_code_list = []
for x in code_list:
# Holds each flattened featured to be concatenated after the loop
assert x.shape[
0] == 1, "can only compress one image at a time, for now"
flat_code_list.append(x.long().flatten())
full_code = torch.cat(flat_code_list) + self.min_val
# use RangeEncoder to write compressed representation to file
encoder = rc.RangeEncoder(savepath)
code_list = full_code.tolist()
encoder.encode(code_list, self.cdf)
encoder.close()
def __init__(self, method):
if not isinstance(method, str):
raise ValueError(f'Invalid method type "{type(method)}"')
from compressai import available_entropy_coders
if method not in available_entropy_coders():
methods = ', '.join(available_entropy_coders())
raise ValueError(f'Unknown entropy coder "{method}"'
f' (available: {methods})')
if method == 'ans':
from compressai import ans # pylint: disable=E0611
encoder = ans.RansEncoder()
decoder = ans.RansDecoder()
elif method == 'rangecoder':
import range_coder # pylint: disable=E0401
encoder = range_coder.RangeEncoder()
decoder = range_coder.RangeDecoder()
self._encoder = encoder
self._decoder = decoder
def apply_range_encoder(seq_data, encodepath, args, config):
resolution = config['resolution']
prob = np.load('data_info/distribution_info_{}.npy'.format(args.model_num))
# Avoid zero prob
modified_freq = prob * resolution + 1
modified_prob = modified_freq / np.sum(modified_freq)
# print(modified_prob)
cum_freq = range_coder.prob_to_cum_freq(modified_prob,
resolution=resolution)
# print('-----')
# print(cum_freq)
# cum_freq = [0] + [i for i in range(1, 256 + 1)]
range_encoder = range_coder.RangeEncoder(encodepath)
# Whether cum_freq resolution influences performance ?
range_encoder.encode(seq_data, cum_freq)
range_encoder.close()
