Get the size of images in the list. Returns: tuple: (width, height) of the images Raises: Assertion

Deserialize member variables from a file into this object

Resize all images with the same parameters. Args: *args, **kwargs: Arguments passed to PIL.Image.resize() Retur

Serialize the object's member variables to a file

Write the buffer content to the logger. Parameters: - buf: The string buffer to write

Clear all keyframes.

Apply a geometric transformation to a list of 3-D points. H: 3x3 or 4x4 projection matrix (typically a Homography) p: numpy/torch/tuple of c

Normalization aligned with DA3's training-time normalize_camera_extrinsics_and_points_batch. Align to first camera + scale by avg point dista

Ensure the HuggingFace repo is available locally, and place the download under your specified checkpoints directory whenever possible. Retur

Apply a geometric transformation to a list of 3-D points. H: 3x3 or 4x4 projection matrix (typically a Homography) p: numpy/torch/tuple of c

pack and write to a binary file. :param fid: :param data: data to send, if multiple elements are sent at the same time, they should be enc

Open an image or a depthmap with opencv-python.

Collect pixel colors from each frame's image according to a mask. Args: images_raw: Tensor (N, 3, H, W) [0, 1] mask_per_frame: (N

Applies radial or OpenCV distortion to the given 2D points. Args: extra_params (torch.Tensor or numpy.ndarray): Distortion parameter

Save point cloud + optional camera frustums to a GLB file. Args: points: (M, 3) world-coordinate points colors: (M, 3) RGB colors

Compute the inverse of each 4x4 (or 3x4) SE3 matrix in a batch. If `R` and `T` are provided, they must correspond to the rotation and transl

Crop all images with the same parameters. Args: *args, **kwargs: Arguments passed to PIL.Image.crop() Returns:

Unproject depth maps into a world-coordinate point cloud. Args: depths: (N, H, W) depth maps extrinsics: (N, 3, 4) cam2world

应用旋转位置编码到输入特征 核心思想：使用复数乘法实现特征旋转，保持相对位置信息 数学原理： 对于2D向量 [x1, x2]，旋转θ角度可以表示为复数乘法： (x1 + ix2) * e^(iθ) = (x1 + ix2) * (cos(

Retrieve image by frame ID, or None if not available.

Convert a pose encoding back to camera extrinsics and intrinsics. This function performs the inverse operation of extri_intri_to_pose_encoding,

Applies radial or OpenCV distortion to the given 2D points. Args: extra_params (torch.Tensor or numpy.ndarray): Distortion parameter

Invert a torch or numpy matrix

Read and unpack the next bytes from a binary file. :param fid: :param num_bytes: Sum of combination of {2, 4, 8}, e.g. 2, 6, 16, 30, etc.

List files in dir_path matching exts, sorted lexicographically by filename.

Store a frame image. Evicts oldest if at capacity.

Create an object from config. The config is expected to contains the following: __object__: path: path.to.module name: MyClas

Create an object from config. The config is expected to contains the following: __object__: path: path.to.module name: MyClas

Forward pass of the GCT model. Args: images: Input images [S, 3, H, W] or [B, S, 3, H, W], in range [0, 1] q

Perform 2D max pooling on the input array. Convenience function for 2D max pooling, commonly used in computer vision and image processin

AUC via histogram bins. Matches DA3 calculate_auc_np.

Precompute RoPE coefficients.

Precompute RoPE coefficients.

Decode patch tokens and heads to get final predictions. Params: - images: input images (B, nimgs, 3, H, W) in [0, 1]

Get the sequence parallel group the caller rank belongs to.

Handle export directory

Process single image

Query scenes by tags. Args: tags: list[str], list of tag values (e.g. ["sparse", "indoor"]) operator: "AND" or "OR"

Unified inference interface

Append confidence and return True if drought is triggered.

Convert input to torch tensor if it's not already one.

Convert input to torch tensor if it's not already one.