I have the similar problem in evaluate.py,the error is：ensorflow.python.framework.errors_impl.InvalidArgumentError: Number of ways to split should evenly divide the split dimension, but got split_dim 2 (size = 1) and num_split 3

[HanChen-HUST]

[[{{node split}} = Split[T=DT_FLOAT, num_split=3, _device="/device:CPU:0"](split/split_dim, ToFloat)]] [{{node IteratorGetNext}} = IteratorGetNextoutput_shapes=[[?,?,?,3],[?,?,?,1]], output_types=[DT_FLOAT，DT_INT32], _device="/job:localhost/replica:0/task:0/device:CPU:0"]] Hope to get your answer,that will be great apperciate

Originally posted by @chenhust1995 in https://github.com/rishizek/tensorflow-deeplab-v3-plus/issues/12#issuecomment-448083099

[DRACOyu]

@chenhust1995 hello, I meet the same question，as follows: Key Variable not found in checkpoint [[Node: save/RestoreV2 = RestoreV2[dtypes=[DT_INT32], _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_save/Const_0_0, save/RestoreV2/tensor_names, save/RestoreV2/shape_and_slices)]] I removed the tf.train.init_checkpoint,this problem still remains, I use inceptionV3 as the baseline, from your answer, checkpoint needs to be updated as well.so how to update the checkpoint? hope your answer,thanks

[HanChen-HUST]

@DRACOyu Hi. Firstly I used coco-stuff dataset to use as my dataset,when programming evalute.py,the above problem occurs,then I changed the dataset from coco-stuff to pascal voc 2012,it still has the same problem,then i found the ground-truth picture in SegmentationClassAug has three channels,so i write a small script to change them to single channels,then it could passed.So i think maybe you could try to set the picture in SegmentationClassAug to singe channel.

[DRACOyu]

@chenhust1995 I used lip dataset that the ground-truth picture has single channel，not this reason, I don't use fitune model, why this question still remain

[HanChen-HUST]

@rishizek