arkaju / u-net-satellite Goto Github PK

Can you please share the data for training and testing?

Not sure if still active.
It might make more sense to update from ipython-sql to Jupysql.
Adding more context on depracation.

This transition should help you avoid compatibility issues and get access to newer features.
Just switching the name and version as backward compatibility is on.

TypeError: Object dtype dtype('O') has no native HDF5 equivalent

"for i, image in enumerate(mask_test):"
1 I have tested the model it is only puting the available mask in the predicted directory
3. How to apply the model on new imagery and save its mask ?
4. The below line of code "mask_pred = model.predict(img_val, verbose=1)" is supposed to showcase the new mask
5. then why "for i, image in enumerate(mask_test):" on test mask is done for writing in predicted folder ?

hello, I try to train the network using your data provided through google colab. However, I trained around 4 hours and go through 70 epochs. The error still around 0.3280 and early stopped. Then i tried to predict but get nothing but only black. Any recommendations on how to deal with the problem ? Please see the attached screenshots.

hello can you give us if you have any trained model that worked on your dataset
I do not have enought gpu to train by myself
thank you

The dataset link given at the end of the readme (https://people.eecs.berkeley.edu/~tinghuiz/projects/pix2pix/datasets/maps.tar.gz)
has each file with image and network as single image.
Can you please share how the same dataset was used in your code?

Thank you