admin/lerobot
1
0
Fork 0
mirror of https://github.com/huggingface/lerobot.git synced 2026-07-23 09:46:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(TIFF): add missing quantization and cleanup for TIFF files

Browse Source
This commit is contained in:
CarolinePascal
2026-05-26 14:52:08 +02:00
parent 621f4191be
commit 3d67784f8c
1 changed files with 18 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/lerobot/datasets/video_utils.py
+18 -5
View File
@@ -444,7 +444,8 @@ def encode_video_frames(
video_path.parent.mkdir(parents=True, exist_ok=True) video_path.parent.mkdir(parents=True, exist_ok=True)
# Get input frames # Get input frames
suffix = ".png" if not isinstance(video_encoder, DepthEncoderConfig) else ".tiff" is_depth = isinstance(video_encoder, DepthEncoderConfig)
suffix = ".png" if not is_depth else ".tiff"
template = "frame-" + ("[0-9]" * 6) + suffix template = "frame-" + ("[0-9]" * 6) + suffix
input_list = sorted( input_list = sorted(
glob.glob(str(imgs_dir / template)), key=lambda x: int(x.split("-")[-1].split(".")[0]) glob.glob(str(imgs_dir / template)), key=lambda x: int(x.split("-")[-1].split(".")[0])
@@ -472,8 +473,19 @@ def encode_video_frames(
# Loop through input frames and encode them # Loop through input frames and encode them
for input_data in input_list: for input_data in input_list:
with Image.open(input_data) as input_image: with Image.open(input_data) as input_image:
input_image = input_image.convert("RGB") if is_depth:
input_frame = av.VideoFrame.from_image(input_image) input_frame = quantize_depth(
np.array(input_image),
depth_min=video_encoder.depth_min,
depth_max=video_encoder.depth_max,
shift=video_encoder.shift,
use_log=video_encoder.use_log,
pix_fmt=video_encoder.pix_fmt,
video_backend="pyav"
)
else:
input_image = input_image.convert("RGB")
input_frame = av.VideoFrame.from_image(input_image)
packet = output_stream.encode(input_frame) packet = output_stream.encode(input_frame)
if packet: if packet:
output.mux(packet) output.mux(packet)
@@ -863,7 +875,7 @@ class StreamingVideoEncoder:
Args: Args:
video_keys: List of video feature keys (e.g. ["observation.images.laptop"]) video_keys: List of video feature keys (e.g. ["observation.images.laptop"])
temp_dir: Base directory for temporary MP4 files temp_dir: Base directory for temporary MP4 files
depth_video_keys: List of video feature keys that carry depth maps (e.g. depth_video_keys: List of video or image feature keys that carry depth maps (e.g.
["observation.images.laptop_depth"]). Defaults to ``[]`` (no depth keys). ["observation.images.laptop_depth"]). Defaults to ``[]`` (no depth keys).
""" """
if self._episode_active: if self._episode_active:
@@ -1207,7 +1219,8 @@ class VideoEncodingManager:
img_dir = self.dataset.root / "images" img_dir = self.dataset.root / "images"
if img_dir.exists(): if img_dir.exists():
png_files = list(img_dir.rglob("*.png")) png_files = list(img_dir.rglob("*.png"))
if len(png_files) == 0: tiff_files = list(img_dir.rglob("*.tiff"))
if len(png_files) == 0 and len(tiff_files) == 0:
shutil.rmtree(img_dir) shutil.rmtree(img_dir)
logger.debug("Cleaned up empty images directory") logger.debug("Cleaned up empty images directory")
else: else: