Inputs: - vecs: Tensor of shape (O, D) giving vectors - boxes: Tensor of shape (O, 4) giving bounding boxes in the format [x0, y0, x1, y1] in the [0, 1] coordinate space - obj_to_img: LongTensor of shape (O,) mapping each element of vecs to an image, where each element is in the ran
(vecs, boxes, obj_to_img, H, W=None, pooling='sum')
source not stored for this graph (policy: none)
no test coverage detected