Class ResNetFPNModel

examples/FasterRCNN/modeling/generalized_rcnn.py:199–325

class ResNetFPNModel(GeneralizedRCNN):

    def inputs(self):
        ret = [
            tf.TensorSpec((None, None, 3), tf.float32, 'image')]
        num_anchors = len(cfg.RPN.ANCHOR_RATIOS)
        for k in range(len(cfg.FPN.ANCHOR_STRIDES)):
            ret.extend([
                tf.TensorSpec((None, None, num_anchors), tf.int32,
                              'anchor_labels_lvl{}'.format(k + 2)),
                tf.TensorSpec((None, None, num_anchors, 4), tf.float32,
                              'anchor_boxes_lvl{}'.format(k + 2))])
        ret.extend([
            tf.TensorSpec((None, 4), tf.float32, 'gt_boxes'),
            tf.TensorSpec((None,), tf.int64, 'gt_labels')])  # all > 0
        if cfg.MODE_MASK:
            ret.append(
                tf.TensorSpec((None, None, None), tf.uint8, 'gt_masks_packed')
            )
        return ret

    def slice_feature_and_anchors(self, p23456, anchors):
        for i in range(len(cfg.FPN.ANCHOR_STRIDES)):
            with tf.name_scope('FPN_slice_lvl{}'.format(i)):
                anchors[i] = anchors[i].narrow_to(p23456[i])

    def backbone(self, image):
        c2345 = resnet_fpn_backbone(image, cfg.BACKBONE.RESNET_NUM_BLOCKS)
        p23456 = fpn_model('fpn', c2345)
        return p23456

    def rpn(self, image, features, inputs):
        assert len(cfg.RPN.ANCHOR_SIZES) == len(cfg.FPN.ANCHOR_STRIDES)

        image_shape2d = tf.shape(image)[2:]     # h,w
        all_anchors_fpn = get_all_anchors_fpn(
            strides=cfg.FPN.ANCHOR_STRIDES,
            sizes=cfg.RPN.ANCHOR_SIZES,
            ratios=cfg.RPN.ANCHOR_RATIOS,
            max_size=cfg.PREPROC.MAX_SIZE)
        multilevel_anchors = [RPNAnchors(
            all_anchors_fpn[i],
            inputs['anchor_labels_lvl{}'.format(i + 2)],
            inputs['anchor_boxes_lvl{}'.format(i + 2)]) for i in range(len(all_anchors_fpn))]
        self.slice_feature_and_anchors(features, multilevel_anchors)

        # Multi-Level RPN Proposals
        rpn_outputs = [rpn_head('rpn', pi, cfg.FPN.NUM_CHANNEL, len(cfg.RPN.ANCHOR_RATIOS))
                       for pi in features]
        multilevel_label_logits = [k[0] for k in rpn_outputs]
        multilevel_box_logits = [k[1] for k in rpn_outputs]
        multilevel_pred_boxes = [anchor.decode_logits(logits)
                                 for anchor, logits in zip(multilevel_anchors, multilevel_box_logits)]

        proposal_boxes, proposal_scores = generate_fpn_proposals(
            multilevel_pred_boxes, multilevel_label_logits, image_shape2d)

        if self.training: