Darkflow: YOLOv3 support

Just tested, not working.
Parsing ./cfg/yolov3.cfg Layer [shortcut] not implemented
Hope it will be added soon!

yes, tested and got same a problem!

If you read the official yolov3-voc.cfg, you will see the CNN is more big, and that is actually related to the code, obviously, the code need to be rewrote for a new CNN.

Thanks for bringing this up.

I'll try to do minor refactoring gradually, so that the code is easier to read & extend for contributors. I admit after a while not looking at it I'm kind of lost.

YOLO v3 savepb
Layer [shortcut] not implemented

Hey @thtrieu,
Is this supported yet?
Or does this still need to be implemented?
Great work on the repo anyway!

Highly expected.

Here, someone has implemented yolov3 in tensorflow.
https://github.com/mystic123/tensorflow-yolo-v3

Hi @LucasMahieu !
This implementation is only for testing, not for trainig.
I've found this: https://github.com/raytroop/YOLOv3_tf but I haven't tried yet

Hi @ambr89,
Thanks for the link, I will try it.
It would be nicer if Yolov3 was supported in darkflow...

I know, at the moment I'm trying to translate from darknet to darflow the missing layer and new loss function, but it isn't easy !!!

Great !
Which layer is not easy ?
You may need some help ?

At the moment, I'm trying to reproduce loss function.
The upsample layer I hope I did well, but while I haven't finisched I don't know if work.

@ambr89 I can help you with the translation , I have already translated the 3 new darknet layers to darkflow (need to look for the project cause it's been a while) , contact me if you want to colaborate !
cheers

@ambr89, @bendidi, @LucasMahieu have you agreed on what is needed to do next? Do you have a list of milestones we could help with?

Hi everybody!
At the moment, I'm using this project
It isn't darkflow but it's quite similar.

Do you think it will be add in darflow ?
Actually, I don't understand why yolov3 could not be supported directly in darkflow ?

Still not working

Layer [shortcut] not implemented

Still not working...

Parsing ./cfg/yolo.cfg
Layer [shortcut] not implemented

Maybe the support of tiny yolo v3 would be a good beginning.
The shortcut layer is not used in Yolov3-tiny.

So, supporting yolov3 tiny is only question of upsampling layer (which can be implemented with a Resize Nearest Neighbor in tensorflow).

@ambr89 did you succeeded to run tiny yolo v3 with darkflow ?

What about this project:
https://github.com/qqwweee/keras-yolo3

It uses Keras with TF backend.
But not sure whether it can be used to train with my own classes.
Lots of work,, it seems :)

Hi, I'm using https://github.com/mystic123/tensorflow-yolo-v3 and I'm getting different output formats. Does anyone know if there's any implementation yet of Darkflow that supports Yolov3?

Thanks for bringing this up.

I'll try to do minor refactoring gradually, so that the code is easier to read & extend for contributors. I admit after a while not looking at it I'm kind of lost.

@thtrieu hi guys,have you refactoring it,does it work for yolov3,i am wondering for it,ths

Here, someone has implemented yolov3 in tensorflow.
https://github.com/mystic123/tensorflow-yolo-v3

@LucasMahieu it's seems not work for AS demo

@ambr89 I can help you with the translation , I have already translated the 3 new darknet layers to darkflow (need to look for the project cause it's been a while) , contact me if you want to colaborate !
cheers

@bendidi i need ,ths!

Maybe the support of tiny yolo v3 would be a good beginning.
The shortcut layer is not used in Yolov3-tiny.

So, supporting yolov3 tiny is only question of upsampling layer (which can be implemented with a Resize Nearest Neighbor in tensorflow).

@ambr89 did you succeeded to run tiny yolo v3 with darkflow ?

Parsing yolov3-tiny.cfg
Layer [yolo] not implemented
no ,it didn't work

For those wanting to convert Yolo-v3 from Darknet to TensorFlow:

Yolo-v3 support was recently added to DW2TF (see this PR).

Worth noting that unlike Darkflow which is also a runtime env for training/inference, DW2TF can only convert a Darknet model & weights to TensorFlow. Running training/inference on it would then be up to the user.

Are people still trying to chip away at v3 support, or has this gone cold?

Last commit was 11 months ago 🤔

@thtrieu

@thtrieu any update on this? is this issue resolved?

Layer [shortcut] not implemented

I still get the same error

Hi everybody!

I've translate yolov3 from .ckpt to .pb and from .pb to .tflite

I've use keras, something like this
https://github.com/qqwweee/keras-yolo3

Hi everybody!

I've translate yolov3 from .ckpt to .pb and from .pb to .tflite

I've use keras, something like this
https://github.com/qqwweee/keras-yolo3

Hi Sir,
Still waiting for that update. Have to use darknet and thats a bit difficult for windows users.

Any update on this? still waiting for v3 support.