:param input_dim: dimension of input to be embedded :param max_freq_log2: log2 of max freq; min freq is 1 by default :param N

Initializes the 2D RoPE module.

Args: img_size (int, tuple): input image size patch_size (int, tuple): patch size in_chans (int): number

Init. Args: features (int): number of features

Init. Args: features (int): number of features

Initialize the TrackHead module. Args: dim_in (int): Input dimension of tokens from the backbone. patch_size

Self attention block

Cross attention block

observed_uv: 1 1 K 2 this is the 2d tracks image: pycolmap.Image object point3D: pycolmap.Point3D object camera: py

Args: backbone: a backbone module, must follow detectron2's backbone interface sem_seg_head: a module that predicts s

Creates the matcher Params: cost_class: This is the relative weight of the classification error in the matching cost

Create the criterion. Parameters: num_classes: number of object categories, omitting the special no-object category ma