|
|
import torch |
|
|
from torch.utils.data import DataLoader, Dataset |
|
|
import cv2 |
|
|
import os |
|
|
import rasterio |
|
|
import numpy as np |
|
|
from pyproj import Transformer |
|
|
from datetime import date |
|
|
|
|
|
class S5P_EEAAirQualityDataset(Dataset): |
|
|
''' |
|
|
1973/494 train/test air quality dataset for NO2 and O3, measure from S5P, label from EEA |
|
|
annual: 1x56x56x1 annual avg |
|
|
seasonal: 4x56x56x1 seasonal avg |
|
|
s5p nodata: -inf |
|
|
label nodata: -3.4e38 # this needs to be masked out for loss and metric calculation |
|
|
''' |
|
|
|
|
|
def __init__(self, root_dir, modality='no2', mode='annual', split='train', meta=False): |
|
|
self.root_dir = root_dir |
|
|
self.mode = mode |
|
|
self.modality = modality |
|
|
|
|
|
if self.mode == 'annual': |
|
|
mode_dir = 's5p_annual' |
|
|
elif self.mode == 'seasonal': |
|
|
mode_dir = 's5p_seasonal' |
|
|
|
|
|
self.img_dir = os.path.join(root_dir, modality, split, mode_dir) |
|
|
self.label_dir = os.path.join(root_dir, modality, split, 'label_annual') |
|
|
|
|
|
self.fnames = sorted(os.listdir(self.label_dir)) |
|
|
|
|
|
self.meta = meta |
|
|
if self.meta: |
|
|
self.reference_date = date(1970, 1, 1) |
|
|
|
|
|
def __len__(self): |
|
|
return len(self.fnames) |
|
|
|
|
|
def __getitem__(self, idx): |
|
|
fname = self.fnames[idx] |
|
|
label_path = os.path.join(self.label_dir, fname) |
|
|
img_path = os.path.join(self.img_dir, fname.replace('.tif', '')) |
|
|
img_fnames = os.listdir(img_path) |
|
|
img_paths = [] |
|
|
for img_fname in img_fnames: |
|
|
img_paths.append(os.path.join(img_path, img_fname)) |
|
|
|
|
|
|
|
|
imgs = [] |
|
|
meta_infos = [] |
|
|
for img_path in img_paths: |
|
|
with rasterio.open(img_path) as src: |
|
|
img = src.read(1) |
|
|
img = cv2.resize(img, (56,56), interpolation=cv2.INTER_CUBIC) |
|
|
img[np.isnan(img)] = 0 |
|
|
img = torch.from_numpy(img).float() |
|
|
img = img.unsqueeze(0) |
|
|
|
|
|
if self.meta: |
|
|
cx,cy = src.xy(src.height // 2, src.width // 2) |
|
|
crs_transformer = Transformer.from_crs(src.crs, 'epsg:4326') |
|
|
lon, lat = crs_transformer.transform(cx,cy) |
|
|
|
|
|
img_fname = os.path.basename(img_path) |
|
|
date_str = img_fname.split('_')[0][:10] |
|
|
date_obj = date(int(date_str[:4]), int(date_str[5:7]), int(date_str[8:10])) |
|
|
delta = (date_obj - self.reference_date).days |
|
|
meta_info = np.array([lon, lat, delta, 0]).astype(np.float32) |
|
|
else: |
|
|
meta_info = np.array([np.nan,np.nan,np.nan,np.nan]).astype(np.float32) |
|
|
|
|
|
imgs.append(img) |
|
|
meta_infos.append(meta_info) |
|
|
if self.mode == 'seasonal': |
|
|
|
|
|
while len(imgs) < 4: |
|
|
imgs.append(img) |
|
|
img_paths.append(img_path) |
|
|
meta_infos.append(meta_info) |
|
|
|
|
|
with rasterio.open(label_path) as src: |
|
|
label = src.read(1) |
|
|
label = cv2.resize(label, (56,56), interpolation=cv2.INTER_CUBIC) |
|
|
label[label<-1e10] = np.nan |
|
|
label[label>1e10] = np.nan |
|
|
|
|
|
label = torch.from_numpy(label.astype('float32')) |
|
|
|
|
|
|
|
|
if self.mode == 'annual': |
|
|
return imgs[0], meta_infos[0], label |
|
|
elif self.mode == 'seasonal': |
|
|
return imgs[0], imgs[1], imgs[2], imgs[3], meta_infos[0], meta_infos[1], meta_infos[2], meta_infos[3], label |
|
|
|
|
|
if __name__ == '__main__': |
|
|
dataset = S5P_EEAAirQualityDataset(root_dir='./airquality_s5p', modality='no2', mode='annual', split='train') |
|
|
dataloader = DataLoader(dataset, batch_size=1, shuffle=False) |
|
|
for i, data in enumerate(dataloader): |
|
|
print(data[0].shape, data[1].shape, data[2].shape, data[3].shape) |
|
|
break |
