calebrob6 / land-cover Goto Github PK

Hello @calebrob6 ,
Thanks for publishing your training code. I was going through your code and noticed something.

The color augmentation function assumes the array to be num_channels x height x width judging by the following line and the entire function code:

But it is called here with the array sized height x width x num_channels here:

Is it really an error or did I miss something?

Hello,
Thank you for providing the code for your land cover classification algorithm and in particular, the ingenious super-resolution loss function. I have tried to downlead Iowa NDR 2009 land cover dataset with 1 m resolution. Unfortunately, I was not able to find any downloadable map/tif file. They only provide interactive webtools (here and here), which are not useful for measuring accuracy and miou. I was wondering if you had a link to download Iowa dataset.

Thanks,

Hi,

I have a question about variance in sr_loss. Could you please elaborate on why variance is calculated this way?

# Mean var of predicted distribution:

        var = K.sum(masked_probs * (1.0 - masked_probs), axis=(1, 2)) / (
            c_mask_size * c_mask_size
        )  # (16x5) / (16,1) --> shape 16x5`

Basically, in the above line, we calculate sigma(x(1-x)) / (n^2). However, since we had calculated mean (mu) in the previous line, we could use mu in the calculation of variance.

Definition of variance is sigma((x-mu)^2) / n. So I was thinking something like below:
mean = K.sum(masked_probs, axis=(1, 2)) / c_mask_size # (16x5) / (16,1) --> shape 16x5
var = K.sum((K.pow(masked_probs - mean, 2) * c_mask), axis=(1, 2)) / ( c_mask_size ) # (16x5) / (16,1) --> shape 16x5
In the above line, * c_mask sets to zero the irrelevant values outside the current class's mask.

So why not stick to the exact variance formula here? Am I missing something?

Thanks,

Hi,
When I ran the code again, a similar problem occurred. I want to know how to solve it, that would be greatly appreciated.
Traceback (most recent call last):
File "C:/Users/57282/Desktop/land-cover-master/run.py", line 203, in
main()
File "C:/Users/57282/Desktop/land-cover-master/run.py", line 145, in main
train.run_experiment()
File "C:\Users\57282\Desktop\land-cover-master\landcover\train_model_landcover.py", line 441, in run_experiment
model = self.get_model()
File "C:\Users\57282\Desktop\land-cover-master\landcover\train_model_landcover.py", line 378, in get_model
model = models.unet(self.input_shape, self.classes, optimizer, self.loss)
File "C:\Users\57282\Desktop\land-cover-master\landcover\models.py", line 115, in unet
return make_model(i, o, optimizer, loss)
File "C:\Users\57282\Desktop\land-cover-master\landcover\models.py", line 168, in make_model
model.compile(
File "E:\Anaconda\envs\land_cover\lib\site-packages\tensorflow\python\keras\engine\training.py", line 565, in compile
self._validate_compile(optimizer, metrics, **kwargs)
File "E:\Anaconda\envs\land_cover\lib\site-packages\tensorflow\python\keras\engine\training.py", line 2626, in _validate_compile
raise ValueError(
ValueError: Variable (<tf.Variable 'conv2d/kernel:0' shape=(3, 3, 4, 64) dtype=float32, numpy=
array([[[[-0.05316697, 0.01695759, 0.06781096, ..., 0.04036037,
-0.02615297, -0.05482 ],
[-0.03512351, 0.00207546, 0.08749197, ..., -0.03793961,
-0.0256898 , -0.02477491],
[-0.02103255, -0.02145103, -0.09544748, ..., 0.08490782,
0.00683988, -0.00993489],
[ 0.07889044, 0.03161094, -0.02841607, ..., 0.07148486,
-0.08556601, 0.01502845]],

    [[ 0.02166805, -0.0465791 ,  0.03786724, ..., -0.05684982,
      -0.02745763, -0.02497146],
     [-0.0321082 , -0.07753698,  0.08001576, ...,  0.03958925,
      -0.01430742,  0.07854632],
     [ 0.03598402, -0.009101  , -0.08477397, ...,  0.04529131,
      -0.0007516 ,  0.07106098],
     [-0.01125589,  0.02705219, -0.02472276, ...,  0.01507821,
      -0.086858