Problem when loading weights

[eiraola]

Hi, I was following this tutorial

https://divamgupta.com/image-segmentation/2019/06/06/deep-learning-semantic-segmentation-keras.html

And everrything is great, but when I try to use:

from keras_segmentation import predict

predict(
checkpoints_path="checkpoints/vgg_unet_1",
inp="dataset_path/images_prepped_test/0016E5_07965.png",
out_fname="output.png"
)

it says:

TypeError: 'module' object is not callable

The same happens with:

from keras_segmentation import predict_multiple

predict_multiple(
checkpoints_path="checkpoints/vgg_unet_1",
inp_dir="dataset_path/images_prepped_test/",
out_dir="outputs/"
)

cannot import name 'predict_multiple'

cannot figure out what the problem is, any clues?

[eiraola]

Sorry it was a clumsy mistake, I forgot to put predict. before both functions :)

[MuhammadAsadJaved]

Sorry it was a clumsy mistake, I forgot to put predict. before both functions :)

How did you resolve this issue? I have same problem.

[MuhammadAsadJaved]

Sorry it was a clumsy mistake, I forgot to put predict. before both functions :)

What do you mean by "put predict before both function" ?

[MuhammadAsadJaved]

I have resolved this problem. we should import like this.
#Predict one picture
from keras_segmentation.predict import predict

predict(
checkpoints_path="/home/lab/image-segmentation-keras/checkpoints/pspnet50_ade20k.h5",
inp="images_test/input_image.jpg",
out_fname="output_image.png"
)

#Predict multiple images in a directory

from keras_segmentation.predict import predict_multiple

predict_multiple(
checkpoints_path="checkpoints/vgg_unet_1",
inp_dir="images_test/",
out_dir="outputs/"
)

[MuhammadAsadJaved]

Still have same issue, failed to resolve this issue

…

On Sat, Nov 30, 2019 at 9:50 PM Akash Kumar ***@***.***> wrote: How did you resolve the problem of Checkpoint not found?? @MuhammadAsadJaved <https://github.com/MuhammadAsadJaved> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#134?email_source=notifications&email_token=AG4GR5GWKO2EHCAUPCZ5LEDQWJVQ7A5CNFSM4JOPY4Y2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEFQHPAY#issuecomment-559970179>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AG4GR5EQENNLQZCDE54FZQLQWJVQ7ANCNFSM4JOPY4YQ> .

[AKASH2907]

You can just do it by the name of json file that is being saved in ur checkpoints directory. I figured it out after posting that's why deleted my comment.
If ur json file is saved as vgg-net_config.json just pass vgg-net and it will automatically pick up the best weight corresponding to that json file.

[MuhammadAsadJaved]

I downloaded pre trained weights manually and it only contains .h5 file. How to get .config file?

…

On Sun, Dec 1, 2019 at 3:18 AM Akash Kumar ***@***.***> wrote: You can just do it by the name of json file that is being saved in ur checkpoints directory. I figured it out after posting that's why deleted my comment. If ur json file is saved as vgg-net_config.json just pass vgg-net and it will automatically pick up the best weight corresponding to that json file. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#134?email_source=notifications&email_token=AG4GR5HYPBXGULIE7RL7TZTQWK4B3A5CNFSM4JOPY4Y2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEFQSCNA#issuecomment-560013620>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AG4GR5DFWR7Y2YWWIJRTODLQWK4B3ANCNFSM4JOPY4YQ> .

[AKASH2907]

No, that's not what you have to do. As you train the model, weights get saved in the folder you specify with the specific name as vgg_unet_1.1, vgg_unet_1.2, etc. for each epoch and a json file named as vgg_unet_1_config.json.
When you predict, give the checkpoints_path+vgg_unet_1 as an argument, then the network will start predicting.

[MuhammadAsadJaved]

Dear i didn’t train the model. I am just using pre trained model from repo and run the example.

…

On Sun, Dec 1, 2019 at 10:00 PM Akash Kumar ***@***.***> wrote: No, that's not what you have to do. As you train the model, weights get saved in the folder you specify with the specific name as vgg_unet_1.1, vgg_unet_1.2, etc. for each epoch and a json file named as vgg_unet_1_config.json. When you predict, give the checkpoints_path+vgg_unet_1 as an argument, then the network will start predicting. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#134?email_source=notifications&email_token=AG4GR5HA3PB6MBDKBS7YUKTQWO7OFA5CNFSM4JOPY4Y2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEFRKA2Q#issuecomment-560111722>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AG4GR5DDUGDXEXO3X4N6ZOTQWO7OFANCNFSM4JOPY4YQ> .

[AKASH2907]

If u dont train it, how r u expecting to get json file and all. What are you trying to give till now at the checkpoints path?? Read the readme file bro.
Checkpoints can't be created without training.

[MuhammadAsadJaved]

Oh ok Thank you. but I was just try to run the demo using pre trained models as mentioned in the readme file. They did not mentioned to generate .json file. so i thought it only .h5 weights and rest of the files are already available.

[AKASH2907]

Nah for running also you need to have some checkpoint file and for that you need to train the model. Vgg_unet will automatically download and upload the weights for vgg16 network

[MuhammadAsadJaved]

OK. Thank you very much.

…

On Mon, Dec 2, 2019 at 8:36 PM Akash Kumar ***@***.***> wrote: Nah for running also you need to have some checkpoint file and for that you need to train the model. Vgg_unet will automatically download and upload the weights for vgg16 network — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#134?email_source=notifications&email_token=AG4GR5CQRGDRQVTLBO3N5CTQWT6M7A5CNFSM4JOPY4Y2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEFTLDSY#issuecomment-560378315>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AG4GR5AFBSD247BYQYAROB3QWT6M7ANCNFSM4JOPY4YQ> .

[aneeqahazmat]

I am unable to read my training data , although training images and maska has data (image) still it gives assertion error.