NMS suppression kernel. Args: iou_mask_ptr: Pointer to precomputed IoU mask, shape: (N, N). keep_mask_ptr: Pointer to keep mask t

Map model-side object index to client-side object id.

Pointer jumping kernel with double buffering to avoid race conditions. Reads from labels_in_ptr and writes to labels_out_ptr.

将相机坐标投影到图像坐标

The command loop for each worker process. It listens to commands from the main process and executes them using the model.

将世界坐标转换为相机坐标

Adds a box prompt and run the inference. The image needs to be set, but not necessarily the text prompt. The box is assumed to be in [

Convert annotation which can be polygons, uncompressed RLE to RLE. Args: ann (dict) : annotation object Returns: ann (rle)

Computes the area of all the boxes. Returns: torch.Tensor: a vector with areas of each box.

Computes area of the mask. Only works with Polygons, using the shoelace formula: https://stackoverflow.com/questions/24467972

Returns the color as a BGR tuple. Returns: Tuple[int, int, int]: BGR tuple. Example: ```

Returns the color as an RGB tuple. Returns: Tuple[int, int, int]: RGB tuple. Example: ```

Optimized implementation of detection <-> track association that minimizes DtoH syncs. Args: det_masks: (N, H, W) tensor of predicte

Camera rotation MCQ: - Keep your verified R_rel and YXZ as_euler. - Choose two most significant axes among yaw/pitch/roll. - De

基于 bbox json 与 mask 路径构造“给定 A 视角的 mask，问物体在 B 视角距离多少”QA。 json_path: 3D bbox 结果文件，单元素格式示例： { "ins_id": "1", "label": "b

基于 bbox json 与 mask 路径构造“给定 A 视角的 mask，问物体在 B 视角的方位”QA。 json_path: 3D bbox 结果文件，单元素格式示例： { "ins_id": "1", "label": "be

Return the color at a given index in the palette. Args: idx (int): Index of the color in the palette. Returns:

Concatenates a list of Keypoints into a single Keypoints Arguments: keypoints_list (list[Keypoints]) Returns:

Concatenates a list of PolygonMasks into a single PolygonMasks Arguments: polymasks_list (list[PolygonMasks]) R

Clip (in place) the boxes by limiting x coordinates to the range [0, width] and y coordinates to the range [0, height]. For

Clone the RotatedBoxes. Returns: RotatedBoxes

Args: rgb (bool): whether to return RGB colors or BGR colors. maximum (int): either 255 or 1 Returns: ndarray: a flo

Returns 'white' or 'black' depending on which color contrasts more with the given RGB value.

Convert COCO format annotations to format expected by render_masklet_frame

创建白色mask图片

Crop each bitmask by the given box, and resize results to (mask_size, mask_size). This can be used to prepare training targets for Ma

Crop each mask by the given box, and resize results to (mask_size, mask_size). This can be used to prepare training targets for Mask

depth: (N, H, W) intrinsics: (N, 3, 3) extrinsics: (N, 4, 4) -- world-to-camera images: (N, H, W, 3) conf: (N, H, W) or None

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

Draws a rectangle on a given PIL image using the provided box coordinates in xywh format. :param image: PIL.Image - The image on which to dra

Draw annotations/segmentaions in Detectron2 Dataset format. Args: dic (dict): annotation/segmentation data of one image,

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

Args: soft_mask (ndarray): float array of shape (H, W), each value in [0, 1]. color: color of the mask. Refer to `mat

从mask_index.json中找到对应的mask_rle

Input: - tgt/tgt_query_pos: nq, bs, d_model -

Input: - tgt: nq, bs, d_model - memory: \\sum{hw}, bs, d_model - pos: \\sum{hw}, bs, d_model