Get the connected components (8-connectivity) of binary masks of shape (N, 1, H, W). Inputs: - mask: A binary mask tensor of shape (N, 1

Abbreviates an integer number with K, M, B, T for thousands, millions, billions and trillions, respectively. Examples: >>> get_hu

Initialize an inference state.

Return true if the current process is the main one

Propagate the input points across frames to track in the entire video.

Resets the image embeddings and other state variables.

Remove all input points or mask in all frames throughout the video.

Args: img_size (int): Input image size. Only relevant for rel pos. patch_size (int): Patch size. in_chans

Initialize the neck :param trunk: the backbone :param position_encoding: the positional encoding to use :param d_model: the di

Args: img_size (int): Input image size. Only relevant for rel pos. patch_size (int): Patch size. in_chans

Remove the non-conditioning memory around the input frame. When users provide correction clicks, the surrounding frames' non-conditio

Remove the non-conditioning memory around the input frame. When users provide correction clicks, the surrounding frames' non-conditio

Extracts memory features from the given features and masks. Args: features (dict): A dictionary containing feature maps

Forward SAM prompt encoders and mask heads. Inputs: - backbone_features: image features of [B, C, H, W] shape - poin

Forward SAM prompt encoders and mask heads. Inputs: - backbone_features: image features of [B, C, H, W] shape - poin

Return a process group based on gloo backend, containing all the ranks The result is cached.

`maskmem_pos_enc` is the same across frames and objects, so we cache it as a constant in the inference session to reduce session stor

`maskmem_pos_enc` is the same across frames and objects, so we cache it as a constant in the inference session to reduce session stor

Positionally encode points that are normalized to [0,1].

Positionally encode points that are normalized to [0,1].

Perform post-processing on output masks.

Predict masks for the given input prompts, using the currently set image. Input prompts are batched torch tensors and are expected to

Reset all tracking inputs and results across the videos.

Reset all tracking inputs and results across the videos.

Add new mask to a frame.

Run all_gather on arbitrary picklable data (not necessarily tensors), similar to `all_gather` above, but using filesystem instead of collecti

Calculates boxes in XYXY format around masks. Return [0,0,0,0] for an empty mask. For input shape C1xC2x...xHxW, the output shape is C1xC2x..

Computes the stability score for a batch of masks. The stability score is the IoU between the binary masks obtained by thresholding the p

Add new points and labels to previous point inputs (add at the end).

A post processor to fill small holes in mask scores with area under `max_area`.

A post processor to fill small holes in mask scores with area under `max_area`.

Get the image feature on the input batch.

Positionally encode points that are not normalized to [0,1].

Positionally encode points that are not normalized to [0,1].

Get 1D sine positional embedding as in the original Transformer paper.

Get the distributed and local rank of the current gpu.

Load a PNG file as a mask and its palette.

Load masks from a directory as a dict of per-object masks.

compute bounding box given an input mask Inputs: - masks: [B, 1, H, W] masks, dtype=torch.Tensor Returns: - box_coords: [B, 1,

Encodes masks to an uncompressed RLE, in the format expected by pycoco tools.

Combine per-object masks into a single mask.

Removes small disconnected regions and holes in a mask. Returns the mask and an indicator of if the mask has been modified.

Resets the image embeddings and other state variables.

Compute a binary mask from an uncompressed RLE.

Save a mask as a PNG file with the given palette.