tf.image.generate_bounding_box_proposals

Generate bounding box proposals from encoded bounding boxes.

tf.image.generate_bounding_box_proposals(
    scores,
    bbox_deltas,
    image_info,
    anchors,
    nms_threshold=0.7,
    pre_nms_topn=6000,
    min_size=16,
    post_nms_topn=300,
    name=None
)

scoresA 4-D float Tensor of shape [num_images, height, width, num_achors] containing scores of the boxes for given anchors, can be unsorted.
bbox_deltasA 4-D float Tensor of shape [num_images, height, width, 4 x num_anchors] encoding boxes with respect to each anchor. Coordinates are given in the form [dy, dx, dh, dw].
image_infoA 2-D float Tensor of shape [num_images, 5] containing image information Height, Width, Scale.
anchorsA 2-D float Tensor of shape [num_anchors, 4] describing the anchor boxes. Boxes are formatted in the form [y1, x1, y2, x2].
nms_thresholdA scalar float Tensor for non-maximal-suppression threshold. Defaults to 0.7.
pre_nms_topnA scalar int Tensor for the number of top scoring boxes to be used as input. Defaults to 6000.
min_sizeA scalar float Tensor. Any box that has a smaller size than min_size will be discarded. Defaults to 16.
post_nms_topnAn integer. Maximum number of rois in the output.
nameA name for this operation (optional).

roisRegion of interest boxes sorted by their scores.
roi_probabilitiesscores of the ROI boxes in the ROIs' Tensor.