Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted
18 Views

Need hekp with openvino_training_extensions (first time use)

have downloaded all openvino_training_extensions from Git

make all steps

by using

python ../../external/models/research/object_detection/model_main.py \ --model_dir=./model \ --pipeline_config_path=./configs/pipeline.config system makes ./model folder but without eval_0/ subfolder. There is why by using tensorboard --logdir=./model it sees no photos.

please, correct my steps

thank you

Vasili

0 Kudos
4 Replies
Highlighted
Employee
18 Views

Dear timofeev, vasil

I assume you are talking about openvino_training_extensions ? Each model has a unique set of steps. For instance Pytorch Action Recognition looks like this.  Tensorflow SSD Detector looks like this . From your above command, I'm not sure what you're trying to do.

May I ask why you are thinking of openvino_training_extensions ? 

Thanks,

Shubha

0 Kudos
Highlighted
Employee
18 Views

Dear timofeev, vasil,

If you have further questions it's best to post your question on github issues for training_extensions as the person who maintains it is quite responsive.

Thanks,

Shubha

0 Kudos
Highlighted
18 Views

I followed  Tensorflow personal_vehicule_bike  steps

after

python ./tools/create_crossroad_extra_tf_records.py \
    --train_image_dir=../../data/airport/train/ \
    --val_image_dir=../../data/airport/val/ \
    --train_annotations_file=../../data/airport/annotation_example_train.json \
    --val_annotations_file=../../data/airport/annotation_example_val.json \
    --output_dir=../../data/airport/tfrecords

i get tfrecords folder as it must be

after that i use

python ../../external/models/research/object_detection/model_main.py \
  --model_dir=./model \
  --pipeline_config_path=./configs/pipeline.config

and i get a /model folder, but in the model folder itselves there are no /eval_0 subfolder with "artifacts"

the "model" folder contains only

-rw-rw-r-- 1 imt imt       81 авг 20 22:18 checkpoint
-rw-rw-r-- 1 imt imt 23293218 авг 20 22:18 events.out.tfevents.1566328688.imt-pc
-rw-rw-r-- 1 imt imt 13164544 авг 20 22:18 graph.pbtxt
-rw-rw-r-- 1 imt imt 22992520 авг 20 22:18 model.ckpt-0.data-00000-of-00001
-rw-rw-r-- 1 imt imt    32164 авг 20 22:18 model.ckpt-0.index
-rw-rw-r-- 1 imt imt  6377804 авг 20 22:18 model.ckpt-0.meta

and by using

tensorboard --logdir=./model

i don't see like this right tensorboard image

no "image" menu

Thank You

Vasili

0 Kudos
Highlighted
Employee
18 Views

Dear timofeev, vasil,

As I said before, if you have further questions it's best to post your question on  https://github.com/opencv/openvino_training_extensions/issues  as the person who maintains it is quite responsive.

For your particular issue though, it doesn't seem like an OpenVino issue or even a training_extensions issue. This blog article how to inspect a pre-trained Tensorflow Model should help. I have used the techniques suggested in this blog article myself to visualize a pre-trained Tensorflow *.pb.

Thanks,

Shubha

0 Kudos