def setup_cfg(args: Args):
# load config from file and command-line arguments
cfg = get_cfg()
add_deeplab_config(cfg)
add_mask_former_config(cfg)
cfg.merge_from_file(args.config_file)
cfg.freeze()
return cfg
def add_panoptic_deeplab_config(cfg):
"""
Add config for Panoptic-DeepLab.
"""
# Reuse DeepLab config.
add_deeplab_config(cfg)
# Target generation parameters.
cfg.INPUT.GAUSSIAN_SIGMA = 10
cfg.INPUT.IGNORE_STUFF_IN_OFFSET = True
cfg.INPUT.SMALL_INSTANCE_AREA = 4096
cfg.INPUT.SMALL_INSTANCE_WEIGHT = 3
cfg.INPUT.IGNORE_CROWD_IN_SEMANTIC = False
# Optimizer type.
cfg.SOLVER.OPTIMIZER = "ADAM"
# Panoptic-DeepLab semantic segmentation head.
# We add an extra convolution before predictor.
cfg.MODEL.SEM_SEG_HEAD.HEAD_CHANNELS = 256
cfg.MODEL.SEM_SEG_HEAD.LOSS_TOP_K = 0.2
# Panoptic-DeepLab instance segmentation head.
cfg.MODEL.INS_EMBED_HEAD = CN()
cfg.MODEL.INS_EMBED_HEAD.NAME = "PanopticDeepLabInsEmbedHead"
cfg.MODEL.INS_EMBED_HEAD.IN_FEATURES = ["res2", "res3", "res5"]
cfg.MODEL.INS_EMBED_HEAD.PROJECT_FEATURES = ["res2", "res3"]
cfg.MODEL.INS_EMBED_HEAD.PROJECT_CHANNELS = [32, 64]
cfg.MODEL.INS_EMBED_HEAD.ASPP_CHANNELS = 256
cfg.MODEL.INS_EMBED_HEAD.ASPP_DILATIONS = [6, 12, 18]
cfg.MODEL.INS_EMBED_HEAD.ASPP_DROPOUT = 0.1
# We add an extra convolution before predictor.
cfg.MODEL.INS_EMBED_HEAD.HEAD_CHANNELS = 32
cfg.MODEL.INS_EMBED_HEAD.CONVS_DIM = 128
cfg.MODEL.INS_EMBED_HEAD.COMMON_STRIDE = 4
cfg.MODEL.INS_EMBED_HEAD.NORM = "SyncBN"
cfg.MODEL.INS_EMBED_HEAD.CENTER_LOSS_WEIGHT = 200.0
cfg.MODEL.INS_EMBED_HEAD.OFFSET_LOSS_WEIGHT = 0.01
# Panoptic-DeepLab post-processing setting.
cfg.MODEL.PANOPTIC_DEEPLAB = CN()
# Stuff area limit, ignore stuff region below this number.
cfg.MODEL.PANOPTIC_DEEPLAB.STUFF_AREA = 2048
cfg.MODEL.PANOPTIC_DEEPLAB.CENTER_THRESHOLD = 0.1
cfg.MODEL.PANOPTIC_DEEPLAB.NMS_KERNEL = 7
cfg.MODEL.PANOPTIC_DEEPLAB.TOP_K_INSTANCE = 200
# If set to False, Panoptic-DeepLab will not evaluate instance segmentation.
cfg.MODEL.PANOPTIC_DEEPLAB.PREDICT_INSTANCES = True
cfg.MODEL.PANOPTIC_DEEPLAB.USE_DEPTHWISE_SEPARABLE_CONV = False
# This is the padding parameter for images with various sizes. ASPP layers
# requires input images to be divisible by the average pooling size and we
# can use `MODEL.PANOPTIC_DEEPLAB.SIZE_DIVISIBILITY` to pad all images to
# a fixed resolution (e.g. 640x640 for COCO) to avoid having a image size
# that is not divisible by ASPP average pooling size.
cfg.MODEL.PANOPTIC_DEEPLAB.SIZE_DIVISIBILITY = -1
# Only evaluates network speed (ignores post-processing).
cfg.MODEL.PANOPTIC_DEEPLAB.BENCHMARK_NETWORK_SPEED = False
# Evaluation batch size:
cfg.DATALOADER.EVAL_BATCH_SIZE = 1
def setup(args):
global ONE_EPOCH
#set the number of GPUs
os.environ["CUDA_VISIBLE_DEVICES"] = "0,1"
# REGISTER DATASETS
dataset_directory = "/home/ndserv05/Documents/Data/Tower_foreground"
# register training and validation datasets with detectron
for d in ["train", "val"]:
# get_darwin_dataset(dataset_directory, d)
DatasetCatalog.register(
"tower_foreground_" + d,
lambda d=d: get_darwin_dataset(dataset_directory, d, categories))
MetadataCatalog.get("tower_foreground_" +
d).set(thing_classes=categories)
# number of epochs to train
EPOCHS = 60
NUM_GPU = 2
# get size of train and val datasets
TRAIN_SIZE = len(DatasetCatalog.get("tower_foreground_train"))
VAL_SIZE = len(DatasetCatalog.get("tower_foreground_val"))
# CONFIGURATION
cfg = get_cfg()
add_deeplab_config(cfg)
cfg.merge_from_file(
"./projects/DeepLab/configs/Cityscapes-SemanticSegmentation/deeplab_v3_plus_R_103_os16_mg124_poly_90k_bs16.yaml"
)
cfg.OUTPUT_DIR = "./output/" + "Tower_foreground" + "{:%Y%m%dT%H%M}".format(
datetime.datetime.now())
# cfg.INPUT.MASK_FORMAT = "bitmask"
cfg.DATASETS.TRAIN = ("tower_foreground_train", )
cfg.DATASETS.TEST = ()
# cfg.TEST.EVAL_PERIOD = 30 # eval period should be one epoch, which is the number of images in training set divided by num_gpu*IMS_PER_BATCH
# cfg.DATALOADER.NUM_WORKERS = 2
cfg.MODEL.WEIGHTS = "model_final_a8a355.pkl" # downloaded from https://github.com/facebookresearch/detectron2/tree/master/projects/DeepLab
cfg.SOLVER.IMS_PER_BATCH = 6
ONE_EPOCH = int(TRAIN_SIZE / (NUM_GPU * cfg.SOLVER.IMS_PER_BATCH))
# cfg.SOLVER.MAX_ITER = ONE_EPOCH*EPOCHS
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 2
cfg.MODEL.SEM_SEG_HEAD.NUM_CLASSES = 2
# cfg.INPUT.CROP.ENABLED = False
cfg.INPUT.CROP.SIZE = (512, 512)
# cfg.MODEL.RESNETS.DEPTH = 50
cfg.freeze()
default_setup(cfg, args)
return cfg
def setup(args):
"""
Create configs and perform basic setups.
"""
cfg = get_cfg()
add_deeplab_config(cfg)
cfg.merge_from_file(args.config_file)
cfg.merge_from_list(args.opts)
cfg.freeze()
default_setup(cfg, args)
return cfg
def _get_config(self):
cfg = get_cfg()
add_deeplab_config(cfg)
defaultConfig = optionsHelper.get_hierarchical_value(self.options, ['options', 'model', 'config', 'value', 'id'])
configFile = os.path.join(os.getcwd(), 'ai/models/detectron2/_functional/configs', defaultConfig)
cfg.merge_from_file(configFile)
# disable SyncBatchNorm if not running on distributed system
if comm.get_world_size() <= 1:
cfg.MODEL.RESNETS.NORM = 'BN'
cfg.MODEL.SEM_SEG_HEAD.NORM = 'BN'
return cfg
def add_panoptic_deeplab_config(cfg):
"""
Add config for Panoptic-DeepLab.
"""
# Reuse DeepLab config.
add_deeplab_config(cfg)
# Target generation parameters.
cfg.INPUT.GAUSSIAN_SIGMA = 10
cfg.INPUT.IGNORE_STUFF_IN_OFFSET = True
cfg.INPUT.SMALL_INSTANCE_AREA = 4096
cfg.INPUT.SMALL_INSTANCE_WEIGHT = 3
cfg.INPUT.IGNORE_CROWD_IN_SEMANTIC = False
# Optimizer type.
cfg.SOLVER.OPTIMIZER = "ADAM"
# Panoptic-DeepLab semantic segmentation head.
# We add an extra convolution before predictor.
cfg.MODEL.SEM_SEG_HEAD.HEAD_CHANNELS = 256
cfg.MODEL.SEM_SEG_HEAD.LOSS_TOP_K = 0.2
# Panoptic-DeepLab instance segmentation head.
cfg.MODEL.INS_EMBED_HEAD = CN()
cfg.MODEL.INS_EMBED_HEAD.NAME = "PanopticDeepLabInsEmbedHead"
cfg.MODEL.INS_EMBED_HEAD.IN_FEATURES = ["res2", "res3", "res5"]
cfg.MODEL.INS_EMBED_HEAD.PROJECT_FEATURES = ["res2", "res3"]
cfg.MODEL.INS_EMBED_HEAD.PROJECT_CHANNELS = [32, 64]
cfg.MODEL.INS_EMBED_HEAD.ASPP_CHANNELS = 256
cfg.MODEL.INS_EMBED_HEAD.ASPP_DILATIONS = [6, 12, 18]
cfg.MODEL.INS_EMBED_HEAD.ASPP_DROPOUT = 0.1
# We add an extra convolution before predictor.
cfg.MODEL.INS_EMBED_HEAD.HEAD_CHANNELS = 32
cfg.MODEL.INS_EMBED_HEAD.CONVS_DIM = 128
cfg.MODEL.INS_EMBED_HEAD.COMMON_STRIDE = 4
cfg.MODEL.INS_EMBED_HEAD.NORM = "SyncBN"
cfg.MODEL.INS_EMBED_HEAD.CENTER_LOSS_WEIGHT = 200.0
cfg.MODEL.INS_EMBED_HEAD.OFFSET_LOSS_WEIGHT = 0.01
# Panoptic-DeepLab post-processing setting.
cfg.MODEL.PANOPTIC_DEEPLAB = CN()
# Stuff area limit, ignore stuff region below this number.
cfg.MODEL.PANOPTIC_DEEPLAB.STUFF_AREA = 2048
cfg.MODEL.PANOPTIC_DEEPLAB.CENTER_THRESHOLD = 0.1
cfg.MODEL.PANOPTIC_DEEPLAB.NMS_KERNEL = 7
cfg.MODEL.PANOPTIC_DEEPLAB.TOP_K_INSTANCE = 200
# If set to False, Panoptic-DeepLab will not evaluate instance segmentation.
cfg.MODEL.PANOPTIC_DEEPLAB.PREDICT_INSTANCES = True
cfg.MODEL.PANOPTIC_DEEPLAB.USE_DEPTHWISE_SEPARABLE_CONV = False
def run(self):
# Core function of your process
input = self.getInput(0)
# Get parameters :
param = self.getParam()
if len(input.data["images"]) > 0:
param.cfg["epochs"] = int(param.cfg["maxIter"] * param.cfg["batchSize"] / len(input.data["images"]))
# complete class names if input dataset has no background class
if not (input.has_bckgnd_class):
tmp_dict = {0: "background"}
for k, name in input.data["metadata"]["category_names"].items():
tmp_dict[k + 1] = name
input.data["metadata"]["category_names"] = tmp_dict
input.has_bckgnd_class = True
param.cfg["classes"] = len(input.data["metadata"]["category_names"])
# Call beginTaskRun for initialization
self.beginTaskRun()
if param.cfg["expertModeCfg"] == "":
# Get default config
cfg = get_cfg()
# Add specific deeplab config
add_deeplab_config(cfg)
cfg.merge_from_file(os.path.dirname(os.path.realpath(__file__)) + "/model/configs/deeplab_v3_plus_R_103_os16_mg124_poly_90k_bs16.yaml")
# Generic dataset names that will be used
cfg.DATASETS.TRAIN = ("datasetTrain",)
cfg.DATASETS.TEST = ("datasetTest",)
cfg.SOLVER.MAX_ITER = param.cfg["maxIter"]
cfg.SOLVER.WARMUP_FACTOR = 0.001
cfg.SOLVER.WARMUP_ITERS = param.cfg["maxIter"] // 5
cfg.SOLVER.POLY_LR_FACTOR = 0.9
cfg.SOLVER.POLY_LR_CONSTANT_FACTOR = 0.0
cfg.MODEL.SEM_SEG_HEAD.NUM_CLASSES = param.cfg["classes"]
cfg.SOLVER.BASE_LR = param.cfg["learningRate"]
cfg.MODEL.SEM_SEG_HEAD.ASPP_CHANNELS = 256
cfg.MODEL.SEM_SEG_HEAD.COMMON_STRIDE = 4
cfg.SOLVER.IMS_PER_BATCH = param.cfg["batchSize"]
cfg.DATALOADER.NUM_WORKERS = 0
cfg.INPUT_SIZE = (param.cfg["inputWidth"], param.cfg["inputHeight"])
cfg.TEST.EVAL_PERIOD = param.cfg["evalPeriod"]
cfg.SPLIT_TRAIN_TEST = param.cfg["splitTrainTest"]
cfg.SPLIT_TRAIN_TEST_SEED = -1
cfg.MODEL.BACKBONE.FREEZE_AT = 5
cfg.CLASS_NAMES = [name for k, name in input.data["metadata"]["category_names"].items()]
if param.cfg["earlyStopping"]:
cfg.PATIENCE = param.cfg["patience"]
else:
cfg.PATIENCE = -1
if param.cfg["outputFolder"] == "":
cfg.OUTPUT_DIR = os.path.dirname(os.path.realpath(__file__)) + "/output"
elif os.path.isdir(param.cfg["outputFolder"]):
cfg.OUTPUT_DIR = param.cfg["outputFolder"]
else:
print("Incorrect output folder path")
else:
cfg = None
with open(param.cfg["expertModeCfg"], 'r') as file:
cfg_data = file.read()
cfg = CfgNode.load_cfg(cfg_data)
if cfg is not None:
deeplabutils.register_train_test(input.data["images"], input.data["metadata"],
train_ratio=cfg.SPLIT_TRAIN_TEST / 100,
seed=cfg.SPLIT_TRAIN_TEST_SEED)
os.makedirs(cfg.OUTPUT_DIR, exist_ok=True)
str_datetime = datetime.now().strftime("%d-%m-%YT%Hh%Mm%Ss")
model_folder = cfg.OUTPUT_DIR + os.path.sep + str_datetime
cfg.OUTPUT_DIR = model_folder
if not os.path.isdir(model_folder):
os.mkdir(model_folder)
cfg.OUTPUT_DIR = model_folder
self.trainer = deeplabutils.MyTrainer(cfg, self)
self.trainer.resume_or_load(resume=False)
print("Starting training job...")
launch(self.trainer.train, num_gpus_per_machine=1)
print("Training job finished.")
self.trainer = None
gc.collect()
torch.cuda.empty_cache()
with open(cfg.OUTPUT_DIR+"/Detectron2_DeepLabV3Plus_Train_Config.yaml", 'w') as file:
file.write(cfg.dump())
else:
print("Error : can't load config file "+param.cfg["expertModeCfg"])
# Call endTaskRun to finalize process
self.endTaskRun()
import pickle
import gzip
from tools.darwin import *
import torch
categories = ["Background", "Tower foreground"]
output_dir = "../Results/Results_detectron_tower_foregound"
weights_dir = "./output/Tower_foreground20210701T0654"
dataset_dir = "../../Data/Tower_foreground"
cfg = get_cfg()
add_deeplab_config(cfg)
cfg.merge_from_file("./projects/DeepLab/configs/Cityscapes-SemanticSegmentation/deeplab_v3_plus_R_103_os16_mg124_poly_90k_bs16.yaml")
cfg.MODEL.WEIGHTS = os.path.join(weights_dir, "model_0089999.pth") # path to the model we just trained
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 2
cfg.MODEL.SEM_SEG_HEAD.NUM_CLASSES = 2
cfg.DATASETS.TEST = ("tower_foreground",)
predictor = DefaultPredictor(cfg)
# this is just because the visualiser needs the metadata
DatasetCatalog.register("tower_foreground", lambda: get_darwin_dataset(dataset_dir, 'val', categories))
MetadataCatalog.get("tower_foreground").set(stuff_classes=categories)
MetadataCatalog.get("tower_foreground").set(stuff_colors=[(255,255,0),(0,0,255)])# BGR (not RGB)
# TEST INFERENCE
# dataset_dicts = get_darwin_dataset(dataset_dir, 'val', categories)