In tensor flow implementation, getting dataId object as Blank

TensorFlow General Discussion
,
1

Tensor {
kept: false,
isDisposedInternal: false,
shape: [ 1, 224, 224, 3 ],
dtype: ‘float32’,
size: 150528,
strides: [ 150528, 672, 3 ],
dataId: {},
id: 10870,
rankType: ‘4’,
scopeId: 5397
}

In the above tensor, i am getting dataId: {},. Trying to implement tensorflow js through node js backend i.e.passing the base64 image string from front end to backend. On the backend side, converting base64 string to “Uint8Array” arraybuffer and then passing to browser.frompixel function. It’s returning the tensor properly but i am always getting dataid object blank.

Same implementation when done through only front end i.e.using HTMLvideoelement, tensor is created properly with all the fields values in it. Also prediction seems to be proper.

Please suggest what is the impact of dataid object blank and how it affects the predictions

3

Hi @Pankaj_Mankar ,

The blank dataId object in your Node.js implementation is likely related to backend-specific differences and should not affect your predictions if the tensor’s shape, type, and values are correct. You can proceed with your implementation, but if you face prediction issues, ensure that your tensor is fully initialized and that the backend is properly configured.

Hoping it would help ,

Thank you .