|
|
from torch.utils.data import DataLoader, Dataset |
|
|
import cv2 |
|
|
import os |
|
|
import rasterio |
|
|
import torch |
|
|
import numpy as np |
|
|
from pyproj import Transformer |
|
|
from datetime import date |
|
|
|
|
|
S3_OLCI_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] |
|
|
|
|
|
BIOMASS_MEAN = 93.8317 |
|
|
BIOMASS_STD = 110.5369 |
|
|
|
|
|
|
|
|
class S3OLCI_BiomassDataset(Dataset): |
|
|
''' |
|
|
4000/1000 train/test images 94x94x21 (full dataset is 25K) |
|
|
CCI biomass regression 282x282 |
|
|
nodata: -inf |
|
|
time series: 1-4 images / location |
|
|
|
|
|
''' |
|
|
|
|
|
|
|
|
def __init__(self, root_dir, split='train', mode='static'): |
|
|
self.root_dir = root_dir |
|
|
self.split = split |
|
|
self.mode = mode |
|
|
self.img_dir = os.path.join(root_dir, split, 's3_olci') |
|
|
self.biomass_dir = os.path.join(root_dir, split, 'biomass') |
|
|
|
|
|
self.fnames = os.listdir(self.biomass_dir) |
|
|
|
|
|
if self.mode == 'static': |
|
|
self.static_csv = os.path.join(root_dir, split, 'static_fnames.csv') |
|
|
with open(self.static_csv, 'r') as f: |
|
|
lines = f.readlines() |
|
|
self.static_img = {} |
|
|
for line in lines: |
|
|
dirname = line.strip().split(',')[0] |
|
|
img_fname = line.strip().split(',')[1] |
|
|
self.static_img[dirname] = img_fname |
|
|
|
|
|
|
|
|
|
|
|
def __len__(self): |
|
|
return len(self.fnames) |
|
|
|
|
|
def __getitem__(self, idx): |
|
|
fname = self.fnames[idx] |
|
|
biomass_path = os.path.join(self.biomass_dir, fname) |
|
|
s3_path = os.path.join(self.img_dir, fname.replace('.tif','')) |
|
|
|
|
|
if self.mode == 'static': |
|
|
img_fname = self.static_img[fname.replace('.tif','')] |
|
|
s3_paths = [os.path.join(s3_path, img_fname)] |
|
|
else: |
|
|
img_fnames = os.listdir(s3_path) |
|
|
s3_paths = [] |
|
|
for img_fname in img_fnames: |
|
|
s3_paths.append(os.path.join(s3_path, img_fname)) |
|
|
|
|
|
imgs = [] |
|
|
img_paths = [] |
|
|
meta_infos = [] |
|
|
for img_path in s3_paths: |
|
|
with rasterio.open(img_path) as src: |
|
|
img = src.read() |
|
|
chs = [] |
|
|
for b in range(21): |
|
|
|
|
|
ch = cv2.resize(img[b], (282,282), interpolation=cv2.INTER_CUBIC) |
|
|
chs.append(ch) |
|
|
img = np.stack(chs) |
|
|
img[np.isnan(img)] = 0 |
|
|
for b in range(21): |
|
|
img[b] = img[b]*S3_OLCI_SCALE[b] |
|
|
img = torch.from_numpy(img).float() |
|
|
|
|
|
if self.meta: |
|
|
cx,cy = src.xy(src.height // 2, src.width // 2) |
|
|
|
|
|
|
|
|
lon, lat = cx, cy |
|
|
img_fname = os.path.basename(img_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, np.nan]).astype(np.float32) |
|
|
else: |
|
|
meta_info = np.array([np.nan,np.nan,np.nan,np.nan]).astype(np.float32) |
|
|
|
|
|
|
|
|
imgs.append(img) |
|
|
img_paths.append(img_path) |
|
|
|
|
|
if self.mode == 'series': |
|
|
|
|
|
while len(imgs) < 4: |
|
|
imgs.append(img) |
|
|
img_paths.append(img_path) |
|
|
meta_infos.append(meta_info) |
|
|
|
|
|
with rasterio.open(biomass_path) as src: |
|
|
biomass = src.read(1) |
|
|
biomass = cv2.resize(biomass, (282,282), interpolation=cv2.INTER_CUBIC) |
|
|
biomass = torch.from_numpy(biomass.astype('float32')) |
|
|
biomass = (biomass - BIOMASS_MEAN) / BIOMASS_STD |
|
|
if self.mode == 'static': |
|
|
return imgs[0], meta_infos[0], biomass |
|
|
elif self.mode == 'series': |
|
|
return imgs[0], imgs[1], imgs[2], imgs[3], meta_infos[0], meta_infos[1], meta_infos[2], meta_infos[3], biomass |
