import kornia as K import torch from torchgeo.datasets.geo import NonGeoDataset import os from collections.abc import Callable, Sequence from torch import Tensor import numpy as np import rasterio import cv2 from pyproj import Transformer from datetime import date from typing import TypeAlias, ClassVar import pathlib import logging logging.getLogger("rasterio").setLevel(logging.ERROR) Path: TypeAlias = str | os.PathLike[str] class SenBenchCloudS3(NonGeoDataset): url = None #base_dir = 'all_imgs' splits = ('train', 'val', 'test') split_filenames = { 'train': 'train.csv', 'val': 'val.csv', 'test': 'test.csv', } all_band_names = ( 'Oa01_radiance', 'Oa02_radiance', 'Oa03_radiance', 'Oa04_radiance', 'Oa05_radiance', 'Oa06_radiance', 'Oa07_radiance', 'Oa08_radiance', 'Oa09_radiance', 'Oa10_radiance', 'Oa11_radiance', 'Oa12_radiance', 'Oa13_radiance', 'Oa14_radiance', 'Oa15_radiance', 'Oa16_radiance', 'Oa17_radiance', 'Oa18_radiance', 'Oa19_radiance', 'Oa20_radiance', 'Oa21_radiance', ) all_band_scale = ( 0.0139465,0.0133873,0.0121481,0.0115198,0.0100953,0.0123538,0.00879161, 0.00876539,0.0095103,0.00773378,0.00675523,0.0071996,0.00749684,0.0086512, 0.00526779,0.00530267,0.00493004,0.00549962,0.00502847,0.00326378,0.00324118) rgb_bands = ('Oa08_radiance', 'Oa06_radiance', 'Oa04_radiance') Cls_index_binary = { 'invalid': 0, # --> 255 should be ignored during training 'clear': 1, # --> 0 'cloud': 2, # --> 1 } Cls_index_multi = { 'invalid': 0, # --> 255 should be ignored during training 'clear': 1, # --> 0 'cloud-sure': 2, # --> 1 'cloud-ambiguous': 3, # --> 2 'cloud shadow': 4, # --> 3 'snow and ice': 5, # --> 4 } def __init__( self, root: Path = 'data', split: str = 'train', bands: Sequence[str] = all_band_names, mode = 'multi', transforms: Callable[[dict[str, Tensor]], dict[str, Tensor]] | None = None, download: bool = False, ) -> None: self.root = root self.transforms = transforms self.download = download #self.checksum = checksum assert split in ['train', 'val', 'test'] self.bands = bands self.band_indices = [(self.all_band_names.index(b)+1) for b in bands if b in self.all_band_names] self.mode = mode self.img_dir = os.path.join(self.root, 's3_olci') self.label_dir = os.path.join(self.root, 'cloud_'+mode) self.split_csv = os.path.join(self.root, self.split_filenames[split]) self.fnames = [] with open(self.split_csv, 'r') as f: lines = f.readlines() for line in lines: fname = line.strip() self.fnames.append(fname) self.reference_date = date(1970, 1, 1) self.patch_area = (8*300/1000)**2 # patchsize 8 pix, gsd 300m def __len__(self): return len(self.fnames) def __getitem__(self, index): images, meta_infos = self._load_image(index) #meta_info = np.array([coord[0], coord[1], np.nan, self.patch_area]).astype(np.float32) label = self._load_target(index) sample = {'image': images, 'mask': label, 'meta': meta_infos} if self.transforms is not None: sample = self.transforms(sample) return sample def _load_image(self, index): fname = self.fnames[index] s3_path = os.path.join(self.img_dir, fname) with rasterio.open(s3_path) as src: img = src.read() img[np.isnan(img)] = 0 chs = [] for b in range(21): ch = img[b]*self.all_band_scale[b] #ch = cv2.resize(ch, (256,256), interpolation=cv2.INTER_CUBIC) chs.append(ch) img = np.stack(chs) img = torch.from_numpy(img).float() # get lon, lat cx,cy = src.xy(src.height // 2, src.width // 2) if src.crs.to_string() != 'EPSG:4326': # convert to lon, lat crs_transformer = Transformer.from_crs(src.crs, 'epsg:4326', always_xy=True) lon, lat = crs_transformer.transform(cx,cy) else: lon, lat = cx, cy # get time img_fname = os.path.basename(s3_path) date_str = img_fname.split('____')[1][:8] date_obj = date(int(date_str[:4]), int(date_str[4:6]), int(date_str[6:8])) delta = (date_obj - self.reference_date).days meta_info = np.array([lon, lat, delta, self.patch_area]).astype(np.float32) meta_info = torch.from_numpy(meta_info) return img, meta_info def _load_target(self, index): fname = self.fnames[index] label_path = os.path.join(self.label_dir, fname) with rasterio.open(label_path) as src: label = src.read(1) #label = cv2.resize(label, (256,256), interpolation=cv2.INTER_NEAREST) # 0-650 label[label==0] = 256 label = label - 1 labels = torch.from_numpy(label).long() return labels class SegDataAugmentation(torch.nn.Module): def __init__(self, split, size): super().__init__() mean = torch.Tensor([0.0]) std = torch.Tensor([1.0]) self.norm = K.augmentation.Normalize(mean=mean, std=std) if split == "train": self.transform = K.augmentation.AugmentationSequential( K.augmentation.Resize(size=size, align_corners=True), K.augmentation.RandomRotation(degrees=90, p=0.5, align_corners=True), K.augmentation.RandomHorizontalFlip(p=0.5), K.augmentation.RandomVerticalFlip(p=0.5), data_keys=["input", "mask"], ) else: self.transform = K.augmentation.AugmentationSequential( K.augmentation.Resize(size=size, align_corners=True), data_keys=["input", "mask"], ) @torch.no_grad() def forward(self, batch: dict[str,]): """Torchgeo returns a dictionary with 'image' and 'label' keys, but engine expects a tuple""" x,mask = batch["image"], batch["mask"] x = self.norm(x) x_out, mask_out = self.transform(x, mask) return x_out.squeeze(0), mask_out.squeeze(0).squeeze(0), batch["meta"] class SenBenchCloudS3Dataset: def __init__(self, config): self.dataset_config = config self.img_size = (config.image_resolution, config.image_resolution) self.root_dir = config.data_path self.bands = config.band_names self.mode = config.mode def create_dataset(self): train_transform = SegDataAugmentation(split="train", size=self.img_size) eval_transform = SegDataAugmentation(split="test", size=self.img_size) dataset_train = SenBenchCloudS3( root=self.root_dir, split="train", bands=self.bands, mode=self.mode, transforms=train_transform ) dataset_val = SenBenchCloudS3( root=self.root_dir, split="val", bands=self.bands, mode=self.mode, transforms=eval_transform ) dataset_test = SenBenchCloudS3( root=self.root_dir, split="test", bands=self.bands, mode=self.mode, transforms=eval_transform ) return dataset_train, dataset_val, dataset_test