Get a dummy object pointer based on an empty mask on the current frame.

Compute stability scores of the mask logits based on the IoU between upper and lower thresholds, similar to https://github.com/fairin

Identifies grid cubes that intersect with the underlying surface by checking if the signs at all corners are not identical.

Identifies grid cubes that intersect with the underlying surface by checking if the signs at all corners are not identical.

Identifies grid edges that intersect with the underlying surface by checking for opposite signs. As each edge can be shared by multi

Identifies grid edges that intersect with the underlying surface by checking for opposite signs. As each edge can be shared by multi

Inference interface for the model Args: x (torch.Tensor): input tensor of shape (b, c, h, w) Returns:

Interpolate colors using trilinear interpolation Args: grid_coords: (N, 3) integer grid coordinates local_coo

Lazy loading of the BiRefNet model

Naive implementation of scaled dot product attention.

Naive implementation of scaled dot product attention.

Normalizes the given weights to be non-negative. If input weights are None, it creates and returns a set of weights of ones.

Normalizes the given weights to be non-negative. If input weights are None, it creates and returns a set of weights of ones.

Fuse the current frame's visual feature map with previous memory.

Process frame attention blocks. We keep tokens in shape (B*S, P, C).

Process frame attention blocks. We keep tokens in shape (B*S, P, C).

Process global attention blocks. We keep tokens in shape (B, S*P, C).

Process global attention blocks. We keep tokens in shape (B, S*P, C).

Reset all tracking inputs and results across the videos.

Performs feature rotation by splitting and recombining feature dimensions. Args: x: Input tensor to rotate. Returns:

Run inference with HF SAM

Run the memory encoder on `high_res_masks`. This is usually after applying non-overlapping constraints to object scores. Since their

Run inference with original SAM

Create sinusoidal position embeddings. Args: x: a 1-D Tensor of N indices Returns: an (N, D) Tensor

Create sinusoidal position embeddings. Args: x: a 1-D Tensor of N indices Returns: an (N, D) Tensor

Returns torch.sqrt(torch.max(0, x)) but with a zero subgradient where x is 0.

Returns torch.sqrt(torch.max(0, x)) but with a zero subgradient where x is 0.

Connects four neighboring dual vertices to form a quadrilateral. The quadrilaterals are then split into triangles based on the gamma

Connects four neighboring dual vertices to form a quadrilateral. The quadrilaterals are then split into triangles based on the gamma

Directly turn binary `mask_inputs` into a output mask logits without using SAM. (same input and output shapes as in _forward_sam_head

Whether to use multimask output in the SAM head.

Process network output to extract 3D points and confidence values. Args: out: Network output tensor (B, C, H, W) activation:

Process network output to extract 3D points and confidence values. Args: out: Network output tensor (B, C, H, W) activation:

Activate pose parameters with specified activation functions. Args: pred_pose_enc: Tensor containing encoded pose parameters [transl

Add new mask to a frame.

Solve `min sum_i w_i * (a * x_i + b - y_i ) ^ 2`, where `a` and `b` are scalars, with respect to `a` and `b` using least squares. ### Parame