Step 4 of Figure 5 in seglink paper: Rotate the cropped horizontal bbox back to its original direction Args: center: the cen

Step 2 of Figure 5 in seglink paper Rotate bbox horizontally along a `center` point Args: center: the center of rotation

Re-convert to original image distribution, and convert to int if necessary. Returns: Centered image.

Online Hard Negative Mining. scores: the scores of being predicted as negative cls n_pos: the number of positive sampl

backbone net of vgg16

Filter out bounding boxes based on (relative )overlap with reference box [0, 0, 1, 1]. Remove completely bounding boxes, or assign negative l

Matching a collection of detected boxes with groundtruth values. Does not accept batched-inputs. The algorithm goes as follows: for every dete

Resize bounding boxes based on a reference bounding box, assuming that the latter is [0, 0, 1, 1] after transform. Useful for updating a colle

Compute f-mean with precision and recall

Given a dataset dataset_name and a split_name returns a Dataset. Args: dataset_name: String, the dataset_name of the dataset. spli

Returns preprocessing_fn(image, height, width, **kwargs). Args: name: The name of the preprocessing function. is_training: `True` if

tell if the center of the anchor is in the rect represented using xs and ys and bbox_idx

Match anchors to text boxes. Return: seg_labels: shape = (N,), the seg_labels of segments. each value is the index of matched box i

Re-convert to original image distribution, and convert to int if necessary. Numpy version. Returns: Centered image.

Compute precision and recall from true positives and false positives booleans arrays

Pre-process an given image. Args: image: A `Tensor` representing an image of arbitrary size. output_height: The height of the image a

Random flip left-right of an image and its bounding boxes.

Resize an image and bounding boxes.

Crops and/or pads an image to a target width and height. Resizes an image to a target width and height by either centrally cropping the image

Divides two values, returning 0 if the denominator is <= 0. Args: numerator: A real `Tensor`. denominator: A real `Tensor`, with dtype

Args: seg_scores: the scores of segments being positive link_scores: the scores of linkage being positive seg_offsets_pre

Streaming computation of True and False Positive arrays.

xs, ys: tensors reprensenting ground truth bbox, both with shape=(N, 4), values in 0~1