Compute a matrix of intersection over union values for two lists of bounding boxes using broadcasting :param preds: matrix of predicted bound

Finds matches between prediction and ground truth instances. Returns: gt_match: 1-D array. For each GT box it has the index of the

Return a valid action object with type GO_BACK.

Return a valid action object with type GO_FORWARD.

Return a valid action object with type TYPE.

Return a valid action object with type NEW_TAB.

Return a valid action object with type PAGE_CLOSE.

Return a valid action object with type PAGE_FOCUS.

Optionally query LLM for success.

Checks for episode success, and formulates human feedback if not success User feedback from metadata

Peform crop on the bounding boxes given the offsets. Args: boxes (ndarray or None): bounding boxes to peform crop. The dimension

Peform crop on the bounding boxes given the offsets. Args: boxes (ndarray or None): bounding boxes to peform crop. The dimension

Args: binary_mask (ndarray): numpy array of shape (H, W), where H is the image height and W is the image width. E

Args: binary_mask (ndarray): numpy array of shape (H, W), where H is the image height and W is the image width. E

min_width and min_height: Minimum dimensions of the bounding box to be plotted.

Args: box_coord (tuple): a tuple containing x0, y0, x1, y1 coordinates, where x0 and y0 are the coordinates of th

Draw instance-level prediction results on an image. Args: predictions (Instances): the output of an instance detection/s

Dynamically load the specified class. :param package_path: Path to the package to load :param class_name: Name of the class within the pa

Damerau–Levenshtein edit distance from: https://github.com/gfairchild/pyxDamerauLevenshtein Lowest among K predictions

Damerau–Levenshtein edit distance from: https://github.com/gfairchild/pyxDamerauLevenshtein Lowest among K predictions

apply the VL or language encoder to the inputs

Compute AP/MR for all classes, then averages

Retrieve examples of revising demonstration script command: (str) input command dialogue output is sorted by embedding distan

Click at coordinates (left, top).

This function explores the environment based on

Performs feature extraction

mode: '2D' or '3D' to get positional 3D from position + centroids or 2D encodings

Extract 2D or 3D positional encodings for the image features mode: '2D' or '3D' to get positional 3D from position + centroids or 2D

given a frame number and a starting pt offset, compute the expected pt for the frame. Frame is assumed to be an index (0-based)

given a frame number and a starting pt offset, compute the expected pt for the frame. Frame is assumed to be an index (0-based)

Read txt files containing bounding boxes (ground truth and detections).

modes: reproject_centroid,

Return the space of serialized actions.

Returns a camera matrix from image size and fov.

Get a copy of the default config.

Get formatted dialogue

Return agent's current pose in the form of a Pose object :param agent_id: 0 for Commander and 1 for Driver/ Follower

Loads example embeddings from .npy files for each input example. We store them as torch Tensors (N, D).

Returns: ndarray: the visualized image of shape (H, W, 3) (RGB) in uint8 type. The shape is scale