Convert MP4 to YUV using ffmpeg Args: input_mp4: Input MP4 filename Returns: bool: True if conversion succe

Convert MP4 to YUV using ffmpeg Args: input_mp4: Input MP4 filename Returns: bool: True if conversion succe

Check if the nvcc is avaiable on the machine.

Decode artifacts back into quantized-domain tensors.

Run decompression Args: compress_dir (str): directory that contains compressed files Returns: Dict[str, Tens

Convert depth maps to 3D points Args: depths: Depth maps [..., H, W, 1] camtoworlds: Camera-to-world transformation matrices [...

Encode tensor dict into artifacts on disk and return metadata.

Collect bitrate/metric samples for one method metric, filtering invalid rows. Parameters ---------- variants : OrderedDict[str, dict]

Return the display order of methods exactly as they appear in the JSON. Parameters ---------- results : OrderedDict[str, OrderedDict[str,

Apply Gaussian filtering followed by downsampling to reduce aliasing artifacts. Args: input_tensor (torch.Tensor): Input tensor

Calculate bitrate from compression folder size and frame number.

Calculate bitrate from compression folder size and frame number.

Aggregate sizes (in bytes) for each component in the compression folder. Components include: means, sh0, sh1, sh2, sh3, quats, scales, opacities,

Aggregate sizes (in bytes) for each component in the compression folder. Components include: means, sh0, sh1, sh2, sh3, quats, scales, opacities,

R: c2w T: w2c

R: c2w T: w2c

Return a copy of the quantizer registry mapping name -> class.

Runs multidimensional B-spline interpolation on the input points.

Loads attributes specified by `param_names` from decoded YUV files.

Load overrides from a YAML file.

Load PSNR, SSIM, and LPIPS metrics from a JSON file. Args: json_path: Path to the JSON file containing metrics Retu

Load the dynamic benchmark JSON into a nested OrderedDict structure. Parameters ---------- path : Path Location of the JSON file