[Help needed]input must be 4-dimensional about tfgo HOT 11 CLOSED

@leopku I also encountered the same problem, do you use readImg.ResizeArea(image.Size{Height:224, Width:224})？

Yes.

Thank you, it helped me a lot.
Thanks also to the author

from tfgo.

You need to add the batch dimension to the input tensor, in order to correctly feed the model with a (1, 573,860,3) shaped tensor

You can use tg.Batchify more or less in this way (code not tested I'm not able to run it right now, I'm sorry)

img := tg.Batchify(scope, []tf.Output{image.Read(scope, "/tmp/test.jpeg", 3)})

from tfgo.

@galeone Big Thanks.

This time error was panic: Can not squeeze dim[1], expected a dimension of 1, got 44

core code

		img := tg.Batchify(scope, []tf.Output{image.Read(scope, "/tmp/a72.jpg", 3).Value()})
		input := tg.Exec(scope, []tf.Output{img}, nil, &tf.SessionOptions{})

		results := model.Exec(
			[]tf.Output{model.Op("StatefulPartitionedCall", 0)},
			map[tf.Output]*tf.Tensor{model.Op("serving_default_input", 0): input[0]},
		)
		log.Debug().Interface("results", results[0].Value()).Msg("")

from tfgo.

I don't know from where the 44 came from - is it an input dimension?

Can you show me the shape of img after the batchify call?

from tfgo.

The shape was [1, ?, ?, 3].

Maybe 44 came from input dimension.

I wrote the input dimension into a file. Now I'm analyzing it.

from tfgo.

Finally fixed by comparing codes between which using tensorflow official go package.

from tfgo.

Out of curiosity - can you show me how you fixed it?

from tfgo.

Out of curiosity - can you show me how you fixed it?

I forgot to resize the image 😭 . Resizing the image before batchify can fix it.

This tfgo library is so cool. I really like it.

Finally, thank you for your reply and this amazing library.

from tfgo.

Thank you for the kind words!! 😃

from tfgo.

@leopku I also encountered the same problem, do you use readImg.ResizeArea(image.Size{Height:224, Width:224})？

from tfgo.

@leopku I also encountered the same problem, do you use readImg.ResizeArea(image.Size{Height:224, Width:224})？

Yes.

from tfgo.