moved get_video_duration_in_s to video_utils and replaced subprocess and ffmpeg with pyAV

src/lerobot/datasets/lerobot_dataset.py

@@ -53,7 +53,6 @@ from lerobot.datasets.utils import (
     get_parquet_file_size_in_mb,
     get_parquet_num_frames,
     get_safe_version,
-    get_video_duration_in_s,
     get_video_size_in_mb,
     hf_transform_to_torch,
     is_valid_version,
@@ -77,6 +76,7 @@ from lerobot.datasets.video_utils import (
     decode_video_frames,
     encode_video_frames,
     get_safe_default_codec,
+    get_video_duration_in_s,
     get_video_info,
 )
 

src/lerobot/datasets/utils.py

@@ -17,7 +17,6 @@ import contextlib
 import importlib.resources
 import json
 import logging
-import subprocess
 from collections.abc import Iterator
 from pathlib import Path
 from pprint import pformat
@@ -142,26 +141,6 @@ def get_video_size_in_mb(mp4_path: Path) -> float:
     return file_size_mb
 
 
-def get_video_duration_in_s(mp4_file: Path) -> float:
-    # TODO(rcadene): move to video_utils.py
-    command = [
-        "ffprobe",
-        "-v",
-        "error",
-        "-show_entries",
-        "format=duration",
-        "-of",
-        "default=noprint_wrappers=1:nokey=1",
-        str(mp4_file),
-    ]
-    result = subprocess.run(
-        command,
-        stdout=subprocess.PIPE,
-        stderr=subprocess.STDOUT,
-    )
-    return float(result.stdout)
-
-
 def flatten_dict(d: dict, parent_key: str = "", sep: str = "/") -> dict:
     """Flatten a nested dictionary structure by collapsing nested keys into one key with a separator.
 

src/lerobot/datasets/v30/convert_dataset_v21_to_v30.py

@@ -62,7 +62,6 @@ from lerobot.datasets.utils import (
     flatten_dict,
     get_parquet_file_size_in_mb,
     get_parquet_num_frames,
-    get_video_duration_in_s,
     get_video_size_in_mb,
     load_info,
     update_chunk_file_indices,
@@ -71,7 +70,7 @@ from lerobot.datasets.utils import (
     write_stats,
     write_tasks,
 )
-from lerobot.datasets.video_utils import concat_video_files
+from lerobot.datasets.video_utils import concat_video_files, get_video_duration_in_s
 
 V21 = "v2.1"
 

src/lerobot/datasets/video_utils.py

@@ -267,7 +267,7 @@ def encode_video_frames(
     video_path = Path(video_path)
     imgs_dir = Path(imgs_dir)
 
-    video_path.parent.mkdir(parents=True, exist_ok=overwrite)
+    video_path.parent.mkdir(parents=True, exist_ok=True)
 
     # Encoders/pixel formats incompatibility check
     if (vcodec == "libsvtav1" or vcodec == "hevc") and pix_fmt == "yuv444p":
@@ -277,9 +277,9 @@ def encode_video_frames(
         pix_fmt = "yuv420p"
 
     # Get input frames
-    template = "frame_" + ("[0-9]" * 6) + ".png"
+    template = "frame-" + ("[0-9]" * 6) + ".png"
     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])
     )
 
     # Define video output frame size (assuming all input frames are the same size)
@@ -512,3 +512,25 @@ def get_image_pixel_channels(image: Image):
         return 4  # RGBA
     else:
         raise ValueError("Unknown format")
+
+
+def get_video_duration_in_s(video_path: Path | str) -> float:
+    """
+    Get the duration of a video file in seconds using PyAV.
+
+    Args:
+        video_path: Path to the video file.
+
+    Returns:
+        Duration of the video in seconds.
+    """
+    with av.open(str(video_path)) as container:
+        # Get the first video stream
+        video_stream = container.streams.video[0]
+        # Calculate duration: stream.duration * stream.time_base gives duration in seconds
+        if video_stream.duration is not None:
+            duration = float(video_stream.duration * video_stream.time_base)
+        else:
+            # Fallback to container duration if stream duration is not available
+            duration = float(container.duration / av.time_base)
+    return duration