def image_boxes(
asset_path: str,
tensor_image,
tensor_boxes,
rescale=1,
dataformats="CHW",
asset_rel_path: str = None,
):
if not np:
logger.warning(NUMPY_ERROR_MESSAGE)
return UNKNOWN
tensor_image = to_np(tensor_image)
tensor_image = convert_to_HWC(tensor_image, dataformats)
tensor_boxes = to_np(tensor_boxes)
tensor_image = tensor_image.astype(np.float32) * calculate_scale_factor(
tensor_image
)
return make_image(
asset_path,
tensor_image.astype(np.uint8),
rescale=rescale,
rois=tensor_boxes,
asset_rel_path=asset_rel_path,
)
def histogram(values, bins, max_bins=None):
if not np:
logger.warning(NUMPY_ERROR_MESSAGE)
return UNKNOWN
values = to_np(values).astype(float)
if values.size == 0:
raise ValueError("The input has no element.")
values = values.reshape(-1)
counts, limits = np.histogram(values, bins=bins)
num_bins = len(counts)
if max_bins is not None and num_bins > max_bins:
subsampling = num_bins // max_bins
subsampling_remainder = num_bins % subsampling
if subsampling_remainder != 0:
counts = np.pad(
counts,
pad_width=[[0, subsampling - subsampling_remainder]],
mode="constant",
constant_values=0,
)
counts = counts.reshape(-1, subsampling).sum(axis=-1)
if counts.size == 0:
logger.warning("Tracking an empty histogram")
return UNKNOWN
return V1EventHistogram(values=values, counts=counts)
def metric(value):
if isinstance(value, float):
return value
if not np:
logger.warning(NUMPY_ERROR_MESSAGE)
return UNKNOWN
value = to_np(value)
assert value.squeeze().ndim == 0, "scalar should be 0D"
return float(value)
def video(asset_path, tensor, fps=4, content_type="gif"):
if not np:
logger.warning(NUMPY_ERROR_MESSAGE)
return UNKNOWN
tensor = to_np(tensor)
tensor = prepare_video(tensor)
# If user passes in uint8, then we don't need to rescale by 255
scale_factor = calculate_scale_factor(tensor)
tensor = tensor.astype(np.float32)
tensor = (tensor * scale_factor).astype(np.uint8)
return make_video(asset_path, tensor, fps, content_type)
def image(asset_path, data, rescale=1, dataformats="CHW"):
if not np:
logger.warning(NUMPY_ERROR_MESSAGE)
return UNKNOWN
tensor = to_np(data)
tensor = convert_to_HWC(tensor, dataformats)
# Do not assume that user passes in values in [0, 255], use data type to detect
scale_factor = calculate_scale_factor(tensor)
tensor = tensor.astype(np.float32)
tensor = (tensor * scale_factor).astype(np.uint8)
return make_image(asset_path, tensor, rescale=rescale)
def audio(asset_path: str,
tensor,
sample_rate=44100,
asset_rel_path: str = None):
if not np:
logger.warning(NUMPY_ERROR_MESSAGE)
return UNKNOWN
tensor = to_np(tensor)
tensor = tensor.squeeze()
if abs(tensor).max() > 1:
print("warning: audio amplitude out of range, auto clipped.")
tensor = tensor.clip(-1, 1)
assert tensor.ndim == 1, "input tensor should be 1 dimensional."
tensor_list = [int(32767.0 * x) for x in tensor]
import wave
import struct
check_or_create_path(asset_path, is_dir=False)
wave_write = wave.open(asset_path, "wb")
wave_write.setnchannels(1)
wave_write.setsampwidth(2)
wave_write.setframerate(sample_rate)
tensor_enc = b""
for v in tensor_list:
tensor_enc += struct.pack("<h", v)
wave_write.writeframes(tensor_enc)
wave_write.close()
return V1EventAudio(
sample_rate=sample_rate,
num_channels=1,
length_frames=len(tensor_list),
path=asset_rel_path or asset_path,
content_type="audio/wav",
)
