TensorFlow Compression

Contact owners and managers

1–30 of 127

Welcome to tensorflow-compression.

This group is dedicated to discussing questions and general feedback related to the TensorFlow data compression library. We're discussing problems with training and use of compression models, and any other topics related to machine learning and data compression.

If you think there's a bug or other problem in the library, please head over to github and post an issue there with the information necessary to reproduce the problem.

This group is also where announcements for new releases of the library will be made, so make sure to subscribe to not miss out on new features!

0 selected

Haotian Zhang, Nick Johnston2

Jan 8

general question: any plan to release the models?

We've had a few technical hurdles in getting these models released. Due to the large amount of

unread,

general question: any plan to release the models?

We've had a few technical hurdles in getting these models released. Due to the large amount of

Jan 8

Leonardo Augusto, Fabian Mentzer3

12/16/23

Encode/decode images with higher resolution than the autoencoder input

Ohh, I see. I also found this stackoverflow post that helped me understand the inputs of the layers

unread,

Encode/decode images with higher resolution than the autoencoder input

Ohh, I see. I also found this stackoverflow post that helped me understand the inputs of the layers

12/16/23

10/16/23

Latent space extraction of compression models

Hello team tf-compression, I am conducting a magisterial thesis regarding the study of the effects of

unread,

Latent space extraction of compression models

Hello team tf-compression, I am conducting a magisterial thesis regarding the study of the effects of

10/16/23

Ahmed Ghorbel, … Nikolai Körber4

8/9/23

About testing on different sized images.

Hello, I found my error. While experimenting with different entropy models, I introduced a bug (not

unread,

About testing on different sized images.

Hello, I found my error. While experimenting with different entropy models, I introduced a bug (not

8/9/23

mahadev prasad panda, Johannes Ballé3

2/26/23

NoisyNormalMixture in ContinuousIndexedEntropyModel example

Hi Johannes, Thank you for your reply. Unfortunately, I am not pursuing the project anymore. So I am

unread,

NoisyNormalMixture in ContinuousIndexedEntropyModel example

Hi Johannes, Thank you for your reply. Unfortunately, I am not pursuing the project anymore. So I am

2/26/23

Muhammad Salman Ali, Fabian Mentzer3

12/4/22

Improving the Training Pipeline/Speed for Image Compression

1) The additional gains after three days of training compared to 1 day of training amount to 1.5% to

unread,

Improving the Training Pipeline/Speed for Image Compression

1) The additional gains after three days of training compared to 1 day of training amount to 1.5% to

12/4/22

Su Zi Wei, … Johannes Ballé3

11/14/22

Session crashed after trying to decompress the tfci image in Colab

Hi, I'm not aware of any recent RAM issues. Have you tried going back to the previous TF/TFC

unread,

Session crashed after trying to decompress the tfci image in Colab

Hi, I'm not aware of any recent RAM issues. Have you tried going back to the previous TF/TFC

11/14/22

Johannes Ballé

8/9/22

Re: [tensorflow-compression] GDN Negative Parameters

Hi Scott, the parameters should not become negative, at least in the default implementation. We

unread,

Re: [tensorflow-compression] GDN Negative Parameters

Hi Scott, the parameters should not become negative, at least in the default implementation. We

8/9/22

Scott Ransom, Johannes Ballé3

5/20/22

RDFT of SignalConv2D

Thanks for the explanation :) On Thursday, 19 May 2022 at 09:21:19 UTC+1 Johannes Ballé wrote: Hi

unread,

RDFT of SignalConv2D

Thanks for the explanation :) On Thursday, 19 May 2022 at 09:21:19 UTC+1 Johannes Ballé wrote: Hi

5/20/22

Ashan Perera, … Fabian Mentzer8

5/16/22

HiFiC Target BPP

That's correct, "hific_loss.svg" is the GAN loss. I'll be sure to update after

unread,

HiFiC Target BPP

That's correct, "hific_loss.svg" is the GAN loss. I'll be sure to update after

5/16/22

Linfeng Zheng, Johannes Ballé4

5/13/22

Question about your PCS 2021 paper "3D SCENE COMPRESSION THROUGH ENTROPY PENALIZED NEURAL REPRESENTATION FUNCTIONS"

Yes, that is correct. On Thu, May 12, 2022 at 2:37 AM Linfeng Zheng <zhenglin...@gmail.com

unread,

Question about your PCS 2021 paper "3D SCENE COMPRESSION THROUGH ENTROPY PENALIZED NEURAL REPRESENTATION FUNCTIONS"

Yes, that is correct. On Thu, May 12, 2022 at 2:37 AM Linfeng Zheng <zhenglin...@gmail.com

5/13/22

Sicheng Li, Johannes Ballé2

3/8/22

Any plan on making cNeRF implenmation open source?

Hi, currently we aren't planning to do this, but there is an implementation of the Entropy

unread,

Any plan on making cNeRF implenmation open source?

Hi, currently we aren't planning to do this, but there is an implementation of the Entropy

3/8/22

Priyanka Mudgal, Johannes Ballé2

2/4/22

Prepare .metagraph

+Fabian Mentzer Hi Priyanka, unfortunately I don't know of a good, publicly available way to

unread,

Prepare .metagraph

+Fabian Mentzer Hi Priyanka, unfortunately I don't know of a good, publicly available way to

2/4/22

Mohana, Johannes Ballé3

1/30/22

How is bpp more than packed_bpp?

I checked the math, and this is actually because these bounds only hold in the limit, for an infinite

unread,

How is bpp more than packed_bpp?

I checked the math, and this is actually because these bounds only hold in the limit, for an infinite

1/30/22

Johannes Ballé

1/26/22

Hi everyone, a new release is out: https://github.com/tensorflow/compression/releases/tag/v2.7.0 It

unread,

Hi everyone, a new release is out: https://github.com/tensorflow/compression/releases/tag/v2.7.0 It

1/26/22

Timor Leiderman, Johannes Ballé2

1/8/22

ContinuousBatchedEntropyModel example

Hi Timor, please take a look at this code: https://github.com/tensorflow/compression/blob/master/

unread,

ContinuousBatchedEntropyModel example

Hi Timor, please take a look at this code: https://github.com/tensorflow/compression/blob/master/

1/8/22

Dailan He, David Minnen3

10/27/21

Question about progressive decoding in MS2020 paper

Thanks. It is quite clear to me. 在2021年10月27日星期三 UTC+8 上午8:10:22<David Minnen> 写道： I think the

unread,

Question about progressive decoding in MS2020 paper

Thanks. It is quite clear to me. 在2021年10月27日星期三 UTC+8 上午8:10:22<David Minnen> 写道： I think the

10/27/21

Gabriel Potassi

9/13/21

How can I use a trained model (hific) for decompression?

Hi, tf-compression team Thanks for providing a nice and convenient tool for deep image compression. I

unread,

How can I use a trained model (hific) for decompression?

Hi, tf-compression team Thanks for providing a nice and convenient tool for deep image compression. I

9/13/21

8/29/21

Parameterizing the indexes for an indexed entropy model

Hi there, I'm curious if you have any insight/experience about how a (hyper)decoder network

unread,

Parameterizing the indexes for an indexed entropy model

Hi there, I'm curious if you have any insight/experience about how a (hyper)decoder network

8/29/21

Tomás, Johannes Ballé3

8/27/21

Quite different results with "noisy quantization" v.s. actual quantization

Hi Johannes, Thank you for the great explanation and the paper! It all makes sense now. About finding

unread,

Quite different results with "noisy quantization" v.s. actual quantization

Hi Johannes, Thank you for the great explanation and the paper! It all makes sense now. About finding

8/27/21

8/24/21

About reconstructed image and MS-SSIM for TF2 version model.

Hi Johannes, I trained your model by myself, here I have two queries. Reproducibility Problem: Once

unread,

About reconstructed image and MS-SSIM for TF2 version model.

Hi Johannes, I trained your model by myself, here I have two queries. Reproducibility Problem: Once

8/24/21

diqk...@gmail.com, Johannes Ballé2

8/23/21

About: could not find tensorflow~=2.5.0 while installing tfc_2.2

Hi Khanh, you don't need to install tensorflow_gpu for GPU support, since the plain tensorflow

unread,

About: could not find tensorflow~=2.5.0 while installing tfc_2.2

Hi Khanh, you don't need to install tensorflow_gpu for GPU support, since the plain tensorflow

8/23/21

zhang lu, Johannes Ballé9

8/23/21

About coding with CABAC in compression_Hyperprior

Hi Zhang, most likely you need to change the value of lambda, since it will be different from the MSE

unread,

About coding with CABAC in compression_Hyperprior

Hi Zhang, most likely you need to change the value of lambda, since it will be different from the MSE

8/23/21

zhang lu, Benoit Brummer4

7/28/21

About:the loss function and the value of lambda

And the compressai had loss function of optimized for MS-SSIM,which is loss=λ*(1-D)+R, Is 'D'

unread,

About:the loss function and the value of lambda

And the compressai had loss function of optimized for MS-SSIM,which is loss=λ*(1-D)+R, Is 'D'

7/28/21

Khanh Quoc Dinh, … Wufei Ma10

7/13/21

Implemetation details of paper Scale-Space Flow for Video Compression

Hi Khanh, No problem, I totally understand. Thanks a lot for the results you provided. Those are very

unread,

Implemetation details of paper Scale-Space Flow for Video Compression

Hi Khanh, No problem, I totally understand. Thanks a lot for the results you provided. Those are very

7/13/21

7/1/21

Random crop "End-to-end optimized image compressio

Hi, I'm relatively new to the data augmentation methods.I would like to know more the cropping

unread,

Random crop "End-to-end optimized image compressio

Hi, I'm relatively new to the data augmentation methods.I would like to know more the cropping

7/1/21

Faiz Ullah, Johannes Ballé2

6/18/21

About usage for monochrome Images

Hi Faiz, the pre-trained models were trained for RGB images, but you can feed monochrome images into

unread,

About usage for monochrome Images

Hi Faiz, the pre-trained models were trained for RGB images, but you can feed monochrome images into

6/18/21

Pascal Bacchus, Johannes Ballé2

6/18/21

Performance differences for similar bit rate

Dear Pascal, the results should not differ significantly. You may need to train your models for a

unread,

Performance differences for similar bit rate

Dear Pascal, the results should not differ significantly. You may need to train your models for a

6/18/21

Nick Young, Johannes Ballé2

6/11/21

Problems in installing tfc==1.3 on Linux

Hi, please check that you have a Python version installed that is not newer than 3.7. Back then,

unread,

Problems in installing tfc==1.3 on Linux

Hi, please check that you have a Python version installed that is not newer than 3.7. Back then,

6/11/21

Wang Yan, Johannes Ballé2

6/10/21

Commercial application or products

Dear Yan, as far as I know, most image/video compression in commercial applications is still done

unread,

Commercial application or products

Dear Yan, as far as I know, most image/video compression in commercial applications is still done

6/10/21

Search

Clear search

Close search

Google apps

Main menu