Bad prediction even with high training and validation accuracy

hey guys,

i am using this code for image segmentation:

`from keras_segmentation.models.unet import unet

model = unet(n_classes=3) model.train(n_classes=3, train_images = “/content/Training/images”, train_annotations = “/content/Training/labels”, checkpoints_path = “/content/checkpoints” , epochs=1,validate=True,val_images=“/content/Test/images”,val_annotations=“/content/Test/labels”) ` Epoch 1/1 512/512 [==============================] - 252s 492ms/step - loss: 0.1199 - accuracy: 0.9728 - val_loss: 7.3933e-05 - val_accuracy: 1.0000 saved /content/checkpoints.model.0 Finished Epoch 0

I have high accuracies but bad prediction. I tried different models. All give me the same result.

Image Label Prediction

[reem90]

I had the same issue and I solved it as follows: First of all, the labeled image generated from labelme tool has several channels. While the labeled image used in image-segmentation-keras has one channel. You can check that using Matlab. So, to convert all the labeled images from multiple channels to a single channel, create a small script in Matlab to read the png labeled images and save it again as png. It will directly save it as one channel image. Use these images for training. The generated images will look like black images. just try in a small dataset to make sure that it works.

It worked with me.

Best Regards,

[amil-rp-work]

@TharinduMoraDs
Hey can you please share the python code for the same. I am unfortunately not able to get grayscale image( image of single channel) and with values as 0 or 1.