Train error

[AlessandroZoia-dev]

I executed the train_model.pynb but I have the following error:

InvalidArgumentError Traceback (most recent call last)

in
4 with tf.GradientTape() as tape:
5 rpn_class_loss, rpn_bbox_loss, rcnn_class_loss, rcnn_bbox_loss =
----> 6 model((batch_imgs, batch_metas, batch_bboxes, batch_labels), training=True)
7
8 loss_value = rpn_class_loss + rpn_bbox_loss + rcnn_class_loss + rcnn_bbox_loss

D:\faster2\detection\models\detectors\faster_rcnn.py in call(self, inputs, training)
133 rois, rcnn_labels, rcnn_label_weights, rcnn_delta_targets, rcnn_delta_weights =
134 self.bbox_target.build_targets(
--> 135 proposals, gt_boxes, gt_class_ids, img_metas)
136 else:
137 rois = proposals

D:\faster2\detection\core\bbox\bbox_target.py in build_targets(self, proposals, gt_boxes, gt_class_ids, img_metas)
71 for i in range(batch_size):
72 rois, labels, label_weights, delta_targets, delta_weights = self._build_single_target(
---> 73 proposals[i], gt_boxes[i], gt_class_ids[i], pad_shapes[i], i)
74 rcnn_rois.append(rois)
75 rcnn_labels.append(labels)

D:\faster2\detection\core\bbox\bbox_target.py in _build_single_target(self, proposals, gt_boxes, gt_class_ids, img_shape, batch_ind)
177 tile_delta_weights = tf.zeros((self.num_rcnn_deltas, self.num_classes, 4))
178 ids = tf.stack([tf.range(self.num_rcnn_deltas, dtype=labels.dtype), labels], axis=1)
--> 179 delta_targets = tf.tensor_scatter_nd_update(tile_delta_targets, ids, delta_targets) # [num_rois, self.num_classes, 4]
180 delta_weights = tf.tensor_scatter_nd_update(tile_delta_weights, ids, delta_weights) # [num_rois, self.num_classes, 4]
181

C:\ProgramData\Anaconda3\envs\faster\lib\site-packages\tensorflow_core\python\ops\gen_array_ops.py in tensor_scatter_update(tensor, indices, updates, name)
11081 else:
11082 message = e.message

11083 _six.raise_from(_core._status_to_exception(e.code, message), None)
11084 # Add nodes to the TensorFlow graph.
11085 try:

C:\ProgramData\Anaconda3\envs\faster\lib\site-packages\six.py in raise_from(value, from_value)

InvalidArgumentError: indices[212] = [212, -1] does not index into shape [256,81,4] [Op:TensorScatterUpdate]

Can you help me to understand the problem? Thank you.

[Jackluisus]

I got the same question here