Kernel 'Transform' not registered for backend 'tensorflow'

[adityapatadia]

I am trying to use this repo but getting above error: https://www.npmjs.com/package/@tensorflow-models/face-detection

System information

• TensorFlow.js version (you are using): 3.18
• Are you willing to contribute it (Yes/No): No

There is some issue and this model is unable to run on tfjs-node.

[rthadur]

Transform is not implemented for tfjs-node , this will be a feature request .Any specific reason you are using tfjs-node? Thank you

[google-ml-butler[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you.

[adityapatadia]

Keeping this open.

[dougchestnut]

I just ran into this issue using nodejs. Reason for using tfjs-node, i'm attempting to run face detection on a large collection of images with node. Using node/javascript as it seemed to be the easiest pathway to get started with tensorflow currently. I had no problem using the pre-trained classification and object detection models but ran into this issue with detecting faces.

[adityapatadia]

We are doing inference on server side and I think it’s recommended to use it for speed and efficiency.

On 24-May-2022, at 6:17 AM, Rajeshwar Reddy T @.***> wrote:

 Transform is not implemented for tfjs-node , this will be a feature request .Any specific reason you are using tfjs-node? Thank you

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.

[michaelzangl]

Workaround is to use the CPU backend which is slower but works:

    tf.setBackend('tensorflow'); // < required for decodeImage
    const imageData = await promisify(fs.readFile)(sourceDir + '/' + fileName);
    const image = await tf.node.decodeImage(imageData, 3);
    tf.setBackend('cpu');
    const model = SupportedModels.MediaPipeFaceDetector;
    const detector = await createDetector(model, {runtime: 'tfjs', maxFaces: 1});
    const predictions = await detector.estimateFaces(image);

[adityapatadia]

Is this fixed in newer versions? It seems we are not able to use new face-detection because of this.

[gaikwadrahul8]

Hi, @adityapatadia

Apologize for the delayed response and we are re-visiting our older feature requests and checking whether those feature requests implemented or not as of now and It seems like Face Detection model won't run on tfjs-node you can check package.json file so May I know are you still looking to support this Face Detection model to run on tfjs-node ?

If someone wants to contribute for this feature then you're always welcome and please feel free to do and please refer these links Ref-1, Ref-2 . Thank you!

[adityapatadia]

We will still benefit from this feature and ability to run face detect model on tfjs-node—Regards,Aditya PatadiaOn 09-May-2023, at 2:47 PM, gaikwadrahul8 @.***> wrote: Hi, @adityapatadia Apologize for the delayed response and we are re-visiting our older feature requests and checking whether those feature requests implemented or not as of now and It seems like Face Detection model won't run on tfjs-node you can check package.json file so May I know are you still looking to support this Face Detection model to run on tfjs-node ? If someone wants to contribute for this feature then you're always welcome and please feel free to do and please refer these links Ref-1, Ref-2 . Thank you!

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: @.***>

[adityapatadia]

Hi team, We (and I think anyone running on Nodejs) will benefit from this feature in tfjs-node. Face detection is very widespread use-case and speeding it up is essential for us. Without this, we are forced to use older blazeface repo.

[homerjam]

+1

Reason: using node for face detection

[mprynce]

+1

I'm developing with TFJS for React Native, but learning about TF and writing some test cases with TFJS-node. It would be great to write working code under tfjs-node and then easily move that code over to RN.

[jmgarciasanti]

Same error for me. I'd like to process a "batch" of pictures using NodeJS but "estimatePoses" from posedetection.PoseDetector throws Kernel 'Transform' not registered for backend 'tensorflow'. I'm initializing the PoseDetector with this simple code:

const modelConfig: posedetection.BlazePoseTfjsModelConfig = { modelType: "full", enableSmoothing: true, runtime: "tfjs" };

const model:posedetection.PoseDetector = await posedetection.createDetector(posedetection.SupportedModels.BlazePose, modelConfig);

[Adi1231234]

Hi! I've also encountered the noted error. Are there any updates on resolving this issue, or any recommendations for alternative methods or tools for face detection that are effective with TensorFlow.js in a Node.js environment? Thanks

[alswla]

+1 Hi, I've also encountered the error Kernel 'Transform' not registered for backend 'tensorflow' while using @tensorflow/tfjs-node for face detection.

It would be appreciated if anyone could suggest alternative methods for face detection or provide guidance on how to resolve this issue.

Thanks.

[DmitryFX]

+1

tfjs-node: 4.19.0

tfjs-node: Error: Kernel 'Transform' not registered for backend 'tensorflow'