jacobkoenig / weaklysupervisedcrackseg Goto Github PK

My error is
segmentation.py: error: the following arguments are required: --img_path, --prediction_path, --classifier_type, --classifier_weight_path

I have set
python segmentation.py --img_path=data/images_to_pseudo_label --prediction_path=data/pseudo_labels --classifier_type=R50 --classifier_weight_path=weights_to_classifier.h5

Please help me Thank you

when i set detailed_output=True,
there is a problem:

**cv2.error: OpenCV(4.7.0) 👎 error: (-5:Bad argument) in function 'imwrite'

Overload resolution failed:

• img is not a numerical tuple
• Expected Ptrcv::UMat for argument 'img'**

how can i solve this issue?

First I generated the .h5 file using the following code：

train_dir = 'D:/thesisSupport/code/dataSets/train'

train_generator = train_datagen.flow_from_directory(
    train_dir,
    target_size=(512, 512),
    batch_size=16,
    class_mode='categorical'
)

model = model_factory(classifier_type="R50")
model.compile(optimizer=Adam(lr=0.001), loss='categorical_crossentropy', metrics=['accuracy'])

model.fit(
    train_generator,
    steps_per_epoch=train_generator.samples // 16,
    epochs=3
)

model.save_weights('my_custom_model_weights.h5')

And i set detailed_output = true，modified parameters in return

        if detailed_output:
            # return (
            #     segmentation,
            #     grad_cam_plus,
            #     merged_classifier_pred,
            #     merge_cam_class,
            #     thresholded,
            # )
            return merged_classifier_pred

Finally run the command you provided：python segmentation.py --img_path=D:/thesisSupport/code/dataSets/validation/crack --prediction_path=data/pseudo_labels --classifier_type=R50 --classifier_weight_path=my_custom_model_weights.h5pervisedCrackSeg-master\WeaklySupervisedCrackSeg-master>

Thank you for your dedication, it helped me a lot. But i got the prediciton all black，I don't know which step went wrong.
It would be greatly appreciated if you could reply to this message!!!!
Thanks a lot!!