guicfg.py

from collections import OrderedDict 
classes_dict = OrderedDict({'Squatter':1})
epochs = 300
predict_output_format = 'jpg'   # option include jpg bmp
show_training_page = True

augm_rotation = True
augm_translate = True
augm_flip = True

cls_list = ['Farmland','Trees','Vehicles','Squatter','Solar']

NUM_OF_PRED_PROC = 1

def get(cls_name):
    if cls_name == "Vehicles":
        class cls_cfg:
            my_size = 512
            down_scale = 1
            bands = 4
            #cls_sub_list = OrderedDict({'RedTaxi':1, 'Bus':2, 'Truck':3, 'PrivateCar':4, 'MiniVan':5, 'EngineeringVehicles':6,'Boat':7})
            cls_sub_list = OrderedDict({'Car':1})
            discard_empty = 0.8   # discard some training images without object, 0.8 for dropping 80% 
    elif cls_name == "Trees":
        class cls_cfg:
            my_size = 576
            down_scale = 2
            bands = 5
            cls_sub_list = OrderedDict({'Trees':1})
            discard_empty = 0.8   # discard some training images without object, 0.8 for dropping 80% 
    elif cls_name == 'Squatter':
        class cls_cfg:
            my_size = 512
            down_scale = 4
            bands = 5
            #cls_sub_list = OrderedDict({'Squatter':1,'BuildingTower':1})
            cls_sub_list = OrderedDict({'Squatter':1})
            discard_empty = 0.01   # discard some training images without object, 0.8 for dropping 80% 
    elif cls_name == 'Solar':
        class cls_cfg:
            my_size = 512
            down_scale = 1
            bands = 4
            cls_sub_list = OrderedDict({'Solar':1})
            discard_empty = 0.97   # discard some training images without object, 0.8 for dropping 80% 
    elif cls_name == 'Farmland':
        class cls_cfg:
            my_size = 512
            down_scale = 4
            bands = 4
            cls_sub_list = OrderedDict({'Farmland':1})
            discard_empty = 0.8   # discard some training images without object, 0.8 for dropping 80% 
    elif cls_name == 'House63':
        class cls_cfg:
            my_size = 512
            down_scale = 2
            bands = 3
            cls_sub_list = OrderedDict({'house':1,'garden':2})
            discard_empty = 0.001   # discard some training images without object, 0.8 for dropping 80% 
    return cls_cfg