RandomBatchGeoSampler

[sammyl-99]

Hi,

I am currently running into the issue of using the random batch geo sampler with my custom raster dataset.

when calling the dataloader I am getting the error:
"RTreeError: Coordinates must be in the form (minx, miny, maxx, maxy) or (x, y) for 2D indexes"

I know my dataset has the correct bounding box, as this is printed:
"
nbart Dataset
type: GeoDataset
bbox: BoundingBox(minx=399960.0, maxx=509760.0, miny=6290200.0, maxy=6400000.0, mint=0.0, maxt=9.223372036854776e+18)
size: 1
"

Any help would be appreciated.

[adamjstewart]

Can you share the code or data so I can try to reproduce this?

[sammyl-99]

import matplotlib.pyplot as plt
import os
import torch
from torch.utils.data import DataLoader
from torchgeo.datasets import RasterDataset, stack_samples, unbind_samples
from torchgeo.datasets.utils import download_url
from torchgeo.samplers import RandomGeoSampler
from torchgeo.samplers import RandomBatchGeoSampler
from torchgeo.samplers import BatchGeoSampler

root = "/content/drive/MyDrive/Wentworth/frames"
sorted(os.listdir(root))

class nbart(RasterDataset):
    filename_glob = "T*_complete.tif"
 
    filename_regex = r"^.{6}_(?P<date>\d{8}T\d{6})_complete"
    date_format = "%Y%m%dT%H%M%S"

    is_image = True
    separate_files = False
    all_bands = ["B01", "B02", "B03", "B04","B05","B06","B07","B08", "B09","B10","B11"]
    rgb_bands = ["B04", "B03", "B02"]


    def plot(self, sample):
        # Find the correct band index order
        rgb_indices = []
        for band in self.rgb_bands:
            rgb_indices.append(self.all_bands.index(band))

        # Reorder and rescale the image
        image = sample["image"][rgb_indices].permute(1, 2, 0)
        image = torch.clamp(image / 2000, min=0, max=1)

        # Plot the image
        fig, ax = plt.subplots()
        ax.imshow(image)

        return fig


torch.manual_seed(1)

dataset = nbart(root)
sampler = RandomBatchGeoSampler(dataset, size=32,batch_size=2, length=1000)
dataloader = DataLoader(dataset, sampler=sampler, collate_fn=stack_samples)

for batch in dataloader:

    sample = unbind_samples(batch)[0]
    dataset.plot(sample)
    plt.axis("off")
    plt.show()

[sammyl-99]

the frames are sentinel 2 images which have been resaved using gdal merge to create tiled frames.

[adamjstewart]

This is the same error as #909 with the same solution. You need to use batch_sampler when using RandomBatchGeoSampler.