Calculate Bounding Box Overlaps in Tensorflow -
let's have 2 tensors of bounding box coordinates: n, dimension [n, 4], , k, dimension [k, 4]. each row of each tensor represents x1, y1, x2, , y2 of bounding box.
is there efficient method in tensorflow produce [n, k] matrix m m[i, j] = overlap(n[i, :], k[j, :])? ideally, overlap intersection-over-union, if it's simpler use method work.
the source code of official object detection model has function calculating iou, see function intersection , iou reference
Comments
Post a Comment