File size: 8,352 Bytes
117e206 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 | from dataclasses import InitVar, dataclass, field
import json
from pathlib import Path
import re
from typing import List
import natsort
import numpy as np
import torch
from monai.bundle import ConfigParser
from monai.data import ThreadDataLoader
from surg_seg.Datasets.SegmentationLabelParser import (
LabelInfoReader,
SegmentationLabelInfo,
SegmentationLabelParser,
)
from surg_seg.Datasets.ImageDataset import ImageDirParser, ImageSegmentationDataset
from surg_seg.Datasets.VideoDatasets import CombinedVidDataset
from surg_seg.ImageTransforms.ImageTransforms import ImageTransforms
from surg_seg.Networks.Models import create_FlexibleUnet
from surg_seg.Trainers.Trainer import ModelTrainer
##################################################################
# Concrete implementation of abstract classes
##################################################################
class Ambf5RecSegMapReader(LabelInfoReader):
"""Read the mapping file for ambf multi-class segmentation."""
def __init__(self, mapping_file: Path, annotations_type: str):
"""
Read segmentation labels mapping files
Parameters
----------
mapping_file : Path
annotation_type : str
Either [2colors, 4colors, or 5colors]
"""
super().__init__(mapping_file)
self.annotations_type = annotations_type
self.read()
def read(self):
with open(self.mapping_file, "r") as f:
mapper = json.load(f)
if self.annotations_type in mapper:
mask = mapper[self.annotations_type]
else:
raise RuntimeWarning(
f"annotations type {self.annotations_type} not found in {self.path2mapping}"
)
self.classes_info = [
SegmentationLabelInfo(idx, key, value) for idx, (key, value) in enumerate(mask.items())
]
class Ambf5RecDataReader(ImageDirParser):
def __init__(self, root_dirs: List[Path], annotation_type: str):
"""Image dataset
Parameters
----------
root_dir : Path
annotation_type : str
Either [2colors, 4colors, or 5colors]
"""
super().__init__(root_dirs)
if not isinstance(root_dirs, list):
root_dirs = [root_dirs]
self.image_folder_list = []
for root_dir in root_dirs:
single_folder = SingleFolderReader(root_dir, annotation_type)
self.image_folder_list.append(single_folder)
self.images_list += single_folder.images_path_list
self.labels_list += single_folder.label_path_list
def __len__(self):
return len(self.images_list)
@dataclass
class SingleFolderReader:
"""
Read a single folder of data from the Ambf5Rec dataset
"""
root_dir: Path
annotation_type: InitVar[str]
annotation_path: Path = field(init=False)
image_path_list: List[Path] = field(init=False)
label_path_list: List[Path] = field(init=False)
image_id_list: List[int] = field(init=False)
# Auxiliary variables used to identify duplicated ids in image folder
flag_list: List[int] = field(init=False)
def __post_init__(self, annotation_type):
self.annotation_dir = self.__get_annotation_dir(annotation_type)
self.images_path_list = natsort.natsorted(list((self.root_dir / "raw").glob("*.png")))
self.flag_list = np.zeros(len(self.images_path_list))
self.images_id_list = self.compute_id_list()
self.label_path_list = [self.annotation_dir / img.name for img in self.images_path_list]
def compute_id_list(self):
ids = []
img_name: Path
for img_name in self.images_path_list:
id_match = self.__extract_id(img_name.name)
self.__check_and_mark_id(id_match)
ids.append(id_match)
return ids
def __get_annotation_dir(self, annotation_type):
valid_options = ["2colors", "4colors", "5colors"]
if annotation_type not in valid_options:
raise RuntimeError(
f"{annotation_type} is not a valid annotation.\n Valid annotations are {valid_options}"
)
return self.root_dir / ("annotation" + annotation_type)
def __extract_id(self, img_name: str) -> int:
"""Extract id from image name"""
id_match = re.findall("[0-9]{6}", img_name)
if len(id_match) == 0:
raise RuntimeError(f"Image {img_name} not formatted correctly")
id_match = int(id_match[0])
return id_match
def __check_and_mark_id(self, id_match):
"""Check that there are no duplicated id"""
if self.flag_list[id_match]:
raise RuntimeError(f"Id {id_match} is duplicated")
self.flag_list[id_match] = 1
##################################################################
# Auxiliary functions
##################################################################
##################################################################
# Main functions
##################################################################
def train_with_video_dataset():
device = "cpu"
vid_root = Path("/home/juan1995/research_juan/accelnet_grant/data/rec01/")
vid_filepath = vid_root / "raw/rec01_seg_raw.avi"
seg_filepath = vid_root / "annotation2colors/rec01_seg_annotation2colors.avi"
ds = CombinedVidDataset(vid_filepath, seg_filepath)
dl = ThreadDataLoader(ds, batch_size=4, num_workers=0, shuffle=True)
pretrained_weigths_path = Path("./assets/weights/trained-weights.pt")
model = create_FlexibleUnet(device, pretrained_weigths_path, ds.label_channels)
optimizer = torch.optim.Adam(model.parameters(), 1e-2)
trainer = ModelTrainer(device=device, max_epochs=2)
model, training_stats = trainer.train_model(model, optimizer, dl)
training_stats.plot_stats()
model_path = "./assets/weights/myweights_video"
torch.save(model.state_dict(), model_path)
training_stats.to_pickle(model_path)
def train_with_image_dataset():
# Config parameters
config = ConfigParser()
config.read_config("./training_configs/thin7/ambf_train_config.yaml")
train_config = config.get_parsed_content("ambf_train_config")
train_dir_list = train_config["train_dir_list"]
valid_dir_list = train_config["val_dir_list"]
annotations_type = train_config["annotations_type"]
pretrained_weights_path = train_config["pretrained_weights_path"]
training_output_path = train_config["training_output_path"]
mapping_file = train_config["mapping_file"]
device = train_config["device"]
epochs = train_config["epochs"]
learning_rate = train_config["learning_rate"]
# Train model
train_data_reader = Ambf5RecDataReader(train_dir_list, annotations_type)
valid_data_reader = Ambf5RecDataReader(valid_dir_list, annotations_type)
label_info_reader = Ambf5RecSegMapReader(mapping_file, annotations_type)
label_parser = SegmentationLabelParser(label_info_reader)
ds = ImageSegmentationDataset(
label_parser, train_data_reader, color_transforms=ImageTransforms.img_transforms
)
dl = ThreadDataLoader(ds, batch_size=4, num_workers=0, shuffle=True)
val_ds = ImageSegmentationDataset(
label_parser, valid_data_reader, color_transforms=ImageTransforms.img_transforms
)
val_dl = ThreadDataLoader(val_ds, batch_size=4, num_workers=0, shuffle=True)
print(f"Training dataset size: {len(ds)}")
print(f"Validation dataset size: {len(val_ds)}")
model = create_FlexibleUnet(device, pretrained_weights_path, label_parser.mask_num)
optimizer = torch.optim.Adam(model.parameters(), learning_rate)
trainer = ModelTrainer(device=device, max_epochs=epochs)
model, training_stats = trainer.train_model(model, optimizer, dl, validation_dl=val_dl)
training_output_path.mkdir(exist_ok=True)
torch.save(model.state_dict(), training_output_path / "myweights.pt")
training_stats.to_pickle(training_output_path)
training_stats.plot_stats(file_path=training_output_path)
print(f"Last train IOU {training_stats.iou_list[-1]}")
print(f"Last validation IOU {training_stats.validation_iou_list[-1]}")
def main():
# train_with_video_dataset()
train_with_image_dataset()
if __name__ == "__main__":
main()
|