from spectral import imshow
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
import matplotlib.patches as patches
def make_hparam_string(patch_size):
return "ps%d" % patch_size
# Input data
print("------------------------")
print("Input data")
print("------------------------")
input = IndianPines_Input.IndianPines_Input()
print("Training pixels", np.count_nonzero(input.train_data))
print("Test pixels", np.count_nonzero(input.test_data))
print("------------------------")
# Configurable parameters
config = {}
config['in_channels'] = input.bands
config['num_classes'] = input.num_classes
config['patch_size'] = 5
config['kernel_size'] = 3
config['conv1_channels'] = 32
config['conv2_channels'] = 64
config['fc1_units'] = 1024
config['batch_size'] = 16
from Flevoland import Flevoland_Input
from SanFrancisco import SanFrancisco_Input
from Salinas import Salinas_Input
import time
import numpy as np
from collections import Counter
from spectral import imshow, save_rgb
import CV_Decoder, CV_Postprocessing
import os
import pandas as pd
import CNNTrain_2D
# Input data
images = ["IndianPines", "Pavia", "Flevoland", "SanFrancisco", "Salinas"]
images_inputs = {
"IndianPines": IndianPines_Input.IndianPines_Input(),
"Pavia": Pavia_Input.Pavia_Input(),
"Flevoland": Flevoland_Input.Flevoland_Input(),
"SanFrancisco": SanFrancisco_Input.SanFrancisco_Input(),
"Salinas": Salinas_Input.Salinas_Input()
}
# Select image to test
selected_img = images[3]
input = images_inputs[selected_img]
print("Image:" + selected_img)
for patch_size in [5]:
config = {}