{
	"DETAILS": {
	"NAME": "Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for CoastTrain water/other segmentation of RGB 768x768 orthomosaic images",
    "DATE": "2023-01-27",
    "URL":"10.5281/zenodo.7574784",
    "CITATION":"Buscombe, D. (2022) Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for CoastTrain water/other segmentation of RGB 768x768 orthomosaic images. Zenodo data release 10.5281/zenodo.7574784",
    "QUERIES": "https://github.com/Doodleverse/segmentation_zoo/issues",
    "CREDIT":"Daniel Buscombe, @MARDAScience",
    "INTENDED_PURPOSE":"High-resolution (UAV) image segmentation",
    "KEYWORDS": {
      "1": "image segmentation",
      "2": "The Doodleverse",
      "3": "Residual UNet",
      "4": "aerial",
	  "5": "semantic segmentation",
	  "6": "water",
	  "7": "Seg2Map",
	  "8": "Machine Learning"
    }
	},  
	"DATASET": {
		"NAME": "Coast Train v1 (custom subset)", 
		"SOURCE": "https://doi.org/10.5281/zenodo.6410157",
		"CITATION": "Wernette, P.A., Buscombe, D.D., Favela, J., Fitzpatrick, S., and Goldstein E., 2022, Coast Train--Labeled imagery for training and evaluation of data-driven models for image segmentation: U.S. Geological Survey data release, https://doi.org/10.5066/P91NP87I.",
		"NUMBER_LABELED_IMAGES": 694,
			"CLASSES": {
				"0": "null",
				"1": "water"
			},
			"REMAP_CLASSES": {"0": 0, "1": 0, "2": 1, "3":1, "4":1, "5":1, "6":1,"7":1,"8":1,"9":1,"10":1,"11":1,"12":1},
			"N_DATA_BANDS": 3,
		"BAND_NAMES": {
			"0": "red",
			"1": "green",
			"2": "blue"    }
		},
	"MODEL": {
		"TARGET_SIZE": [768,768],	
	"NAME": "resunet",
    "KERNEL":9,
    "STRIDE":2,
    "FILTERS":6    
	},
	"TRAINING": {
    "BATCH_SIZE": 18,
    "DROPOUT":0.1,
    "DROPOUT_CHANGE_PER_LAYER":0.0,
    "DROPOUT_TYPE":"standard",
    "USE_DROPOUT_ON_UPSAMPLING":false,
    "LOSS":"dice",
	"PATIENCE": 10,
	"MAX_EPOCHS": 100,
	"VALIDATION_SPLIT": 0.5,
	"RAMPUP_EPOCHS": 20,
	"SUSTAIN_EPOCHS": 0.0,
	"EXP_DECAY": 0.9,
	"START_LR":  1e-7,
	"MIN_LR": 1e-7,
	"MAX_LR": 1e-3,
	"MODE": "all"	 
	},
	"AUGMENTATION": {
    "AUGMENTATION_USED": true,
	"AUG_ROT": 5,
	"AUG_ZOOM": 0.05,
	"AUG_WIDTHSHIFT": 0.05,
	"AUG_HEIGHTSHIFT": 0.05,
	"AUG_HFLIP": true,
	"AUG_VFLIP": false,
	"AUG_LOOPS": 10,
	"AUG_COPIES": 5
	}     
}