Webcam analysis: Is it foggy in the mountain?

Keras
,
1

So the idea is easy: caputure a picture of a webcam (in the mountains) and tell if it is foggy or not. (as sometimes it is sunny higher up). I tried different ways, binary and category classification. but it always seems to overfit.

rescale=1./255,
rotation_range=5, #small range because webcam an static picture
width_shift_range=0.1,
height_shift_range=0.1,
shear_range=0.05,
zoom_range=0.05,
horizontal_flip=True,
fill_mode='nearest'

learning_rate 0.001 and 0.0001 gets overfitting
Different batchsizes didn’t help either.

has any one an idea?

3

Hi @Andreas_W, Could you please let us know how many images you are using per class and how many classes you are having in the dataset. Thank You.