如何判断训练数据是有效的

[ainilian]

python train_model.py
报错内容：
InvalidArgumentError Traceback (most recent call last)
in ()
7 # dummpy forward to build network variables
8 # print("batch_imgs {}, batch_metas {}".format(batch_imgs, batch_metas))
----> 9 _ = model((batch_imgs, batch_metas), training=False)

~/software/anaconda3/envs/tf2/lib/python3.6/site-packages/tensorflow_core/python/keras/engine/base_layer.py in call(self, inputs, *args, **kwargs)
820 with base_layer_utils.autocast_context_manager(
821 self._compute_dtype):
--> 822 outputs = self.call(cast_inputs, *args, **kwargs)
823 self._handle_activity_regularization(inputs, outputs)
824 self._set_mask_metadata(inputs, outputs, input_masks)

fasterRCNN/detection/models/detectors/faster_rcnn.py in call(self, inputs, training)
156 print("fasterRcnn rcnn_probs_list {}, rcnn_deltas_list {}, rois_list {}, img_metas {}".format(rcnn_probs_list, rcnn_deltas_list, rois_list, img_metas))
157 detections_list = self.bbox_head.get_bboxes(
--> 158 rcnn_probs_list, rcnn_deltas_list, rois_list, img_metas)
159
160 return detections_list

fasterRCNN/detection/models/bbox_heads/bbox_head.py in get_bboxes(self, rcnn_probs_list, rcnn_deltas_list, rois_list, img_metas)
120 self._get_bboxes_single(
121 rcnn_probs_list[i], rcnn_deltas_list[i], rois_list[i], pad_shapes[i])
--> 122 for i in range(img_metas.shape[0])
123 ]
124 return detections_list

fasterRCNN/detection/models/bbox_heads/bbox_head.py in (.0)
120 self._get_bboxes_single(
121 rcnn_probs_list[i], rcnn_deltas_list[i], rois_list[i], pad_shapes[i])
--> 122 for i in range(img_metas.shape[0])
123 ]
124 return detections_list

fasterRCNN/detection/models/bbox_heads/bbox_head.py in _get_bboxes_single(self, rcnn_probs, rcnn_deltas, rois, img_shape)
202
203 print("nms_kepp {}".format(nms_keep))
--> 204 nms_keep = tf.concat(nms_keep, axis=0)
205
206 # 3. Compute intersection between keep and nms_keep

~/software/anaconda3/envs/tf2/lib/python3.6/site-packages/tensorflow_core/python/util/dispatch.py in wrapper(*args, **kwargs)
178 """Call target, and fall back on dispatchers if there is a TypeError."""
179 try:
--> 180 return target(*args, **kwargs)
181 except (TypeError, ValueError):
182 # Note: convert_to_eager_tensor currently raises a ValueError, not a

~/software/anaconda3/envs/tf2/lib/python3.6/site-packages/tensorflow_core/python/ops/array_ops.py in concat(values, axis, name)
1515 dtype=dtypes.int32).get_shape().assert_has_rank(0)
1516 return identity(values[0], name=name)
-> 1517 return gen_array_ops.concat_v2(values=values, axis=axis, name=name)
1518
1519

~/software/anaconda3/envs/tf2/lib/python3.6/site-packages/tensorflow_core/python/ops/gen_array_ops.py in concat_v2(values, axis, name)
1116 pass # Add nodes to the TensorFlow graph.
1117 except _core._NotOkStatusException as e:
-> 1118 _ops.raise_from_not_ok_status(e, name)
1119 # Add nodes to the TensorFlow graph.
1120 if not isinstance(values, (list, tuple)):

~/software/anaconda3/envs/tf2/lib/python3.6/site-packages/tensorflow_core/python/framework/ops.py in raise_from_not_ok_status(e, name)
6604 message = e.message + (" name: " + name if name is not None else "")
6605 # pylint: disable=protected-access
-> 6606 six.raise_from(core._status_to_exception(e.code, message), None)
6607 # pylint: enable=protected-access
6608

~/software/anaconda3/envs/tf2/lib/python3.6/site-packages/six.py in raise_from(value, from_value)

InvalidArgumentError: OpKernel 'ConcatV2' has constraint on attr 'T' not in NodeDef '[N=0, Tidx=DT_INT32]', KernelDef: 'op: "ConcatV2" device_type: "GPU" constraint { name: "T" allowed_values { list { type: DT_INT32 } } } host_memory_arg: "values" host_memory_arg: "axis" host_memory_arg: "output"' [Op:ConcatV2] name: concat

--------------------------------------------------------------------------------------------------------------# 2. Map over class IDs
nms_keep = []
print("shape {}".format(unique_pre_nms_class_ids.shape)) 打印出的结果：shape (0,)
for i in range(unique_pre_nms_class_ids.shape[0]):
nms_keep.append(nms_keep_map(unique_pre_nms_class_ids[i]))

    print("nms_kepp {}".format(nms_keep))   打印出的结果：nms_kepp []
    nms_keep = tf.concat(nms_keep, axis=0)

---

---

疑问是不是数据出问题了，如果是的话，怎么判断数据是有效的还是无效的

[ainilian]

解决了，换张图片初始化模型就行了

[yunfei1999]

具体换图片的代码在哪里呢？
img, img_meta, bboxes, labels = train_dataset[6] # [N, 4], shape:[N]=data:[62] # 这句吗？
rgb_img = np.round(img + img_mean)
ori_img = get_original_image(img, img_meta, img_mean)

[yunfei1999]

@ainilian