FLAIR#2 Dataset and Datamodule Integration #2394

MathiasBaumgartinger · 2024-11-05T21:53:20Z

FLAIR#2 dataset

The FLAIR #2 <https://github.com/IGNF/FLAIR-2> dataset is an extensive dataset from the French National Institute of Geographical and Forest Information (IGN) that provides a unique and rich resource for large-scale geospatial analysis.
The dataset is sampled countrywide and is composed of over 20 billion annotated pixels of very high resolution aerial imagery at 0.2 m spatial resolution, acquired over three years and different months (spatio-temporal domains).

The FLAIR2 dataset is a dataset for semantic segmentation of aerial images. It contains aerial images, sentinel-2 images and masks for 13 classes.
The dataset is split into a training and test set.

Dataset features:

* over 20 billion annotated pixels
* aerial imagery
    * 5x512x512
    * 0.2m spatial resolution
    * 5 channels (RGB-NIR-Elevation)
* Sentinel-2 imagery
    * 10-20m spatial resolution
    * 10 spectral bands
    * snow/cloud masks (with 0-100 probability)
    * multiple time steps (T)
    * Tx10xWxH, T, W, H are variable
* label (masks)
    * 512x512
    * 13 classes

Dataset classes:

0: "building",
1: "pervious surface",
2: "impervious surface",
3: "bare soil",
4: "water",
5: "coniferous",
6: "deciduous",
7: "brushwood",
8: "vineyard",
9: "herbaceous vegetation",
10: "agricultural land",
11: "plowed land",
12: "other"  

If you use this dataset in your research, please cite the following paper:

* https://doi.org/10.48550/arXiv.2310.13336

Implementation Details

`NonGeoDataset`, `init()`

After discussions following #2303, we decided that at least until faulty mask data are fixed the flair2 ds will be of type NonGeoDataset. Other than with common NonGeoDatasets, FLAIR2 exposes a use_toy and use_sentinel argument. The use_toy-flag will instead use the toy data which is a small subset of data. The use_sentinel argument on the other hand decides whether a sample includes the augmented sentinel data provided by the maintainers of FLAIR2.

`_verify`, `_download`, `_extract`

As each of the splits/sample-types (i.e. [train, test], [aerial, sentinel, labels] are contained in a individual zip download, download and extraction has to happen multiple times. On the other hand, the toy dataset is contained in a singular zip. Furthermore, to map the super-patches of the sentinel data to the actual input image, a flair-2_centroids_sp_to_patch.json is required, which has to be equally has to be downloaded as an individual zip.

`_load_image`, `_load_sentinel`, `_load_target`

For storage reasons, the elevation (5th band) of the image is stored as a uint. The original height thus is multiplied by 5. We decided to divide the height by 5 to get the original height, to make the trained model more usable for other data. See Questions please.

As mentioned previously, additional metadata has to be used to get from the sentinel.npy to the actual area. Initially for debugging reasons, we implemented to return not the cropped image but the original data and the cropping-slices (i.e. indices). Consequently, the images can be plot in a more meaningful matter. Otherwise, the resolution is so low that one can hardly recognize features. This was crucial for debugging to find the correct logic (classic y, x instead of x, y ordering mistake). We do not know if this is smart for "production code". See Questions please.
Moreover, the dimensions of the sentinel data $T \times C=10 \times W \times H$ vary both $T$ and $W$, $H$. This is problematic for the datamodule. We have not done extensive research, but the varying dimensions seem to bug the module. Disabling the use_sentinel-flag will make the module work.

The labels include values from 1 to 19. The datapaper clearly mentions grouping classes $> 13$ into one class other due to underrepresentation. We followed this suggestion. Furthermore, rescaling from 0 to 12 was applied. See Questions please.

Questions

Do you consider the Elevation rescaling as distortion of the dataset? Shall I exclude it? The argument for it would be easier re-usability on new datasets.

For storage optimization reasons, this elevation information is multiplied by a factor of 5 and encoded as a 8bit unsigned integer datatype.

How shall we load/provide sentinel data? As cropped data or any other way. I do not see the current implementation as fit for production.
- Also, how do we want to plot it? The small red rectangle in the example plot above is the actual region. The low resolution is quite observable there.
Shall we rescale the Classes to start from 0? Shall we group the classes as suggested in the datapaper?
Check integrity in download_url does not seem to work (in unit-tests), why?
- I have to call an own check_integrity call otherwise it passes, even if md5s do not match.
The github actions on the forked repo produce a magic ruff error (https://github.com/MathiasBaumgartinger/torchgeo/actions/runs/11687694109/job/32556175383#step:7:1265). Can you help me resolve this mystery?

TODOs/FIXMEs

Extend tests for toy datasets and apply md5 check
Find correct band for plotting sentinel
Datamodule cannot handle sentinel data yet

…mg and msk) Updates in the custom raster dataset tutorial and the actual file documentation. The previous recommended approach (overriding `__get_item__`) is outdated. Refs: microsoft#2292 (reply in thread)

Co-authored-by: Adam J. Stewart <[email protected]>

… type individually

…rrect rng ranges

Not fully functioning yet, contains copy paste from other datasets

Additionally, some small code refactors are done

…d refine plotting

Using the entire sentinel-2 image and a matplotlib patch to debug, otherwise it is really hard to find correct spot due to low resolution

…y()` for sentinel With the nested dict, it was not possible to download dynamically

…sion

md5s might change due to timestamps, this eases the process of changing md5

JacobJeppesen · 2024-11-25T12:08:29Z

torchgeo/datasets/flair2.py

+                to_extract.append(dir_name)
+                continue
+
+            files_glob = os.path.join(downloaded_path, "**", self.globs[train_or_test])


When instantiating the datamodule, it will extract the flair_sen_train zip-file on every run. I traced it here, and I believe it's becuase when going through when dir_name="flair_sen_train", I get files_glob="/home/jhj/dataset_test/FLAIR2/flair_sen_train/**/SEN2_*{0}.npy", with the curly braces at the end. This makes the glob.glob() fail in the next line. I tried manually setting it to files_glob="/home/jhj/dataset_test/FLAIR2/flair_sen_train/**/SEN2_*.npy", which fixed the issue.

Seems like it's the same with flair_2_sen_test, which also gets the {0} in files_glob, and extracts the zip-file on every run.

When instantiating the datamodule, it will extract the flair_sen_train zip-file on every run. I traced it here, and I believe it's becuase when going through when dir_name="flair_sen_train", I get files_glob="/home/jhj/dataset_test/FLAIR2/flair_sen_train/**/SEN2_*{0}.npy", with the curly braces at the end. This makes the glob.glob() fail in the next line. I tried manually setting it to files_glob="/home/jhj/dataset_test/FLAIR2/flair_sen_train/**/SEN2_*.npy", which fixed the issue.

I see why the unexpected behavior appears. For clarification: the format string is necessary, as there are two files inside the directory: mask and data files. To be able to respectively get only the corresponding mask or data file, I had to format the glob.

Would you be willing to share with me your code snippet so I can debug this real quick?

I just manually removed the {0} in the files_glob string before running glob.glob():

files_glob = os.path.join(downloaded_path, "**", self.globs[train_or_test]) if "flair_sen_train" in files_glob: files_glob="/home/jhj/dataset_test/FLAIR2/flair_sen_train/**/SEN2_*.npy" if not glob.glob(files_glob, recursive=True): to_extract.append(dir_name)

I.e., hard-coded files_glob so I could check if it stopped extracting the zip-file on every run. Not really a solution, but just wanted to see if I was the right place in the code.

Just tested it again, and all download and data extraction works now 👍

JacobJeppesen · 2024-11-25T13:24:07Z

Apologize if the comments were a bit scattered, and not really a proper review. I was just trying to run the code, and commented on the go with any issues I encountered 🙂

flair_2_centroids_sp_to_patch.zip vs. flair-2_centroids_sp_to_patch.json Refs: microsoft#2394 (comment)

MathiasBaumgartinger · 2024-11-25T13:31:18Z

I was just testing the pull request, and there was an issue with the download, where it gets a 404 when trying to download https://storage.gra.cloud.ovh.net/v1/AUTH_366279ce616242ebb14161b7991a8461/defi-ia/flair_data_2/flair-2_centroids_sp_to_patch.zip. I think reason stems from inconsistent naming, where the zip file is named flair_2_centroids_sp_to_patch.zip, with underscore after flair, and the json file inside is named flair-2_centroids_sp_to_patch.json, with hyphen after flair. I.e., the correct download url is https://storage.gra.cloud.ovh.net/v1/AUTH_366279ce616242ebb14161b7991a8461/defi-ia/flair_data_2/flair_2_centroids_sp_to_patch.zip.

Great work though! Awesome to see FLAIR being implemented 😃

Weird; I was sure I tried this. Thanks for letting me know.

…extracting Refs: microsoft#2394 (comment)

Instead of loading both sentinel-data/cloudsnowmasks into one sample, store them sperately. Same with the crop_indices.

JacobJeppesen · 2024-11-25T15:04:35Z

I was just testing the pull request, and there was an issue with the download, where it gets a 404 when trying to download https://storage.gra.cloud.ovh.net/v1/AUTH_366279ce616242ebb14161b7991a8461/defi-ia/flair_data_2/flair-2_centroids_sp_to_patch.zip. I think reason stems from inconsistent naming, where the zip file is named flair_2_centroids_sp_to_patch.zip, with underscore after flair, and the json file inside is named flair-2_centroids_sp_to_patch.json, with hyphen after flair. I.e., the correct download url is https://storage.gra.cloud.ovh.net/v1/AUTH_366279ce616242ebb14161b7991a8461/defi-ia/flair_data_2/flair_2_centroids_sp_to_patch.zip.
Great work though! Awesome to see FLAIR being implemented 😃

Weird; I was sure I tried this. Thanks for letting me know.

Yeah, it seems like a weird error. Perhaps they changed it on their side, such that the zip-file naming was consistent. Although the json file inside still uses hyphen instead of underscore 🙂

JacobJeppesen · 2024-11-26T10:22:52Z

torchgeo/datasets/flair2.py

+        files = [
+            dict(image=image, sentinel=sentinel, mask=mask)
+            for image, sentinel, mask in zip(images, sentinels, masks)
+        ]


It seems like this part filters out most of the aerial images. I think it's because each Sentinel-2 image covers multiple aerial images. E.g., flair_aerial_train/D004_2021/Z1_NN contains 100 aerial images, and the corresponding flair_sen_train/D004_2021/Z1_NN contains one .npy file. So the association between Sentinel-2 and the aerial images should probably be done on a per-folder basis instead of per-file basis. E.g., something like:

sentinel_lookup = {"/".join(s["data"].split("/")[-4:-2]): s for s in sentinels} files = [ dict(image=image, sentinel=sentinel_lookup["/".join(image.split("/")[-4:-2])], mask=mask) for image, mask in zip(images, masks) ]

Oh wow! What a severe mistake. Thanks for clarifying. Your approach seems very valid.

That happens sometimes 🙂

I just tried training a model on the aerial data now with the code posted above, and all seems to be working. I haven't looked closely at the Sentinel-2 data, although they looked correct when I sampled a couple of elements in the files list in the debugger.

This bug caused to omit a good 90% of all images. Refs: microsoft#2394 (comment)

JacobJeppesen · 2024-11-26T15:33:30Z

It got all the correct aerial imagery files now and the download and extraction is working. However, I'm getting a much more annoying bug now: IReadBlock failed at X offset ... random error during training.

I've had this issue before with Rasterio, and it was due to concurrency issues. This shouldn't really happen here, as the dataloader just iterates through a list of files, and each sample is a separate tif file. I can see there's also been a discussion here (#594), so perhaps @adamjstewart or @calebrob6 knows what's going on. My own guess is that it could be multiple workers having GDAL do simultaneous scans of the same directory when opening individual files in said directory. I.e., there's often many image files in each sub-folder in the dataset, and GDAL will scan the sub-folder for metadata when opening a file inside it, which could be the culprit. For now at least, I've added the code below in datasets/flair2.py and is testing to see if it solves the issue.

This has been added below the imports in the top:

# Set GDAL to avoid scanning read directories to avoid IReadBlock errors in Rasterio when using num_workers>0 in the 
# dataloader. We should not have any concurrency issues for this dataset, as each worker should read individual tif files,
# but it seems like it might occasionally happen when multiple workers scan the same directory simultaneously. 
os.environ["GDAL_DISABLE_READDIR_ON_OPEN"] = "EMPTY_DIR"

I've also minimized the time the datareader is open in Rasterio by closing it before doing the tensor operations. This shouldn't really solve the issue here, but added it anyways for good practice. In the _load_image() function:

with rasterio.open(path) as f:
    array: np.typing.NDArray[np.int_] = f.read()
tensor = torch.from_numpy(array).float()
if "B05" in self.bands:
    # Height channel will always be the last dimension
    tensor[-1] = torch.div(tensor[-1], 5)

and in the _load_target() function:

with rasterio.open(path) as f:
    array: np.typing.NDArray[np.int_] = f.read(1)
tensor = torch.from_numpy(array).long()
# According to datapaper, the dataset contains classes beyond 13
# however, those are grouped into a single "other" class
# Rescale the classes to be in the range [0, 12] by subtracting 1
torch.clamp(tensor - 1, 0, len(self.classes) - 1, out=tensor)

Not sure if it fixes it, but I'll train a handful of models and try it out. The issue coming from simultaneous directory scans is a bit of a guess.

adamjstewart · 2024-11-26T17:49:18Z

Never had this multiprocessing issue before, and I don't see anything in the code that looks sus. Can you reproduce this issue with the unit tests? If not I can also try downloading the dataset and reproducing locally.

JacobJeppesen · 2024-11-26T20:36:24Z

Unfortunately not. It pops up at random after training for quite a while. Encountered it the first time after having trained for almost 2 full epochs. I.e., it had already read the entire dataset once, and then got IReadBlock error during second epoch. There have been some reports with similar errors here: rasterio/rasterio#2053 . That's generally when reading the same file though, and I agree, nothing looks suspicious in the code. I actually thought my disk was broken, but then I remembered that that was my exact same thought last time I got this error with Rasterio.

I'll try and let it run some more times and see if it keeps happening. Just saw in rasterio/rasterio#2053 (comment) that upgrading to newer version of GDAL might help. I'll also give that a try.

Otherwise the entire file is changed, hence this makes the git-history more traceable

JacobJeppesen · 2024-11-28T13:00:24Z

Finally got it debugged, and it was a system error. Download, extract, and training on the aerial data seems to all work as it should. Sorry about the noise!

Completely unrelated to this PR, but it was a corruption in one of the zpools in ZFS. So if you start using that at some point, and you get weird errors that looks like disk errors, but checking the disk says everything is fine, then try checking the status on your zpools. Spent too much time identifying this 😑

adamjstewart · 2024-11-28T13:12:29Z

Can you resolve the merge conflicts so we can run CI on this PR?

…l and minor bug fixes

…ing inconsistencies in original dataset Naming inconsistencies: `flair-2_centroids_sp_to_patch.json` vs `flair_2_centroids_sp_to_patch.zip`

Refs: 1c2ca19

MathiasBaumgartinger · 2024-11-28T17:10:27Z

I think everything should be on track so far. I am getting a ruff error for unsorted tuples in the [datasets/datamodules]/__init.py__ files. I explicitly tried sorting them using with the vscode sort ascending command (8779f89), but apparently this did work 🤷‍♂️. Let me know if there is anything else I can do for you =)

adamjstewart · 2024-11-28T19:41:24Z

If you use ruff 0.8.0 it will sort __all__ correctly. You can also copy-n-paste the file from the latest version and add the new line.

nilsleh · 2024-11-29T07:03:23Z

torchgeo/datasets/flair2.py

+        """Get statistics (min, max, means, stdvs) for each used band in order.
+
+        Args:
+            split (str): Split for which to get statistics (currently only for train)


For the docstring notion in torchgeo, we do not include the type in the docstring again, only function arguments.

I think this will also resolve the failing docs test

nilsleh · 2024-11-29T07:04:02Z

torchgeo/datasets/flair2.py

+        return tensor
+
+    def _load_sentinel(self, path: Path) -> Tensor:
+        # FIXME: should this really be returned as a tuple?


This can be removed, right?

nilsleh · 2024-11-29T07:04:57Z

torchgeo/datasets/flair2.py

+            self.root,
+            md5=self.md5s.get(url, None) if self.checksum else None,
+        )
+        # FIXME: Why is download_url not checking integrity (tests run through)?


Is this fixed?

This was still an issue last time i checked. As mentioned in the first text of the PRQ, omehow, when I run the pytests with wrong md5 hashes, the integrity is not checked unless I explicitly call it here again (it is implicity called in download_url).

nilsleh · 2024-11-29T07:05:47Z

torchgeo/datasets/flair2.py

+            self.root,
+            md5=self.md5s.get(url, None) if self.checksum else None,
+        )
+        # FIXME: Why is download_url not checking integrity (tests run through)?


how about this FIXME?

I left this here as somewhat of a reminder, because it has the same behavior as the parent class. I.e. currently all tests pass even with wrong md5s.

nilsleh · 2024-11-29T07:06:56Z

torchgeo/datamodules/flair2.py

+        """
+        super().__init__(FLAIR2, batch_size, num_workers, **kwargs)
+
+        self.patch_size = _to_tuple(patch_size)


I think that's a good idea. Could either be included here, or in a separate PR.

nilsleh · 2024-11-29T07:07:46Z

torchgeo/datamodules/flair2.py

@@ -0,0 +1,79 @@
+"""This module contains the FLAIR2DataModule class for loading the FLAIR2 dataset.


I think this also needs the microsoft heading

nilsleh · 2024-11-29T07:09:23Z

tests/data/flair2/FLAIR2/md5s.txt

@@ -0,0 +1,8 @@
+/home/mathias/Dev/forks/torchgeo/tests/data/flair2/FLAIR2/flair_2_labels_test.zip: b13c4a3cb7ebb5cadddc36474bb386f9


maybe remove the personal directory from this text file.

nilsleh · 2024-11-29T07:11:24Z

torchgeo/datasets/flair2.py

+
+        rgb_indices = [self.all_bands.index(band) for band in self.rgb_bands]
+        # Check if RGB bands are present in self.bands
+        if not all([band in self.bands for band in self.rgb_bands]):


The Codecoverage is indicating that the RGB Band Missing is not being hit, so I think you just need to add a separate plot test similar to

torchgeo/tests/datasets/test_eurosat.py

Line 110 in 2f3e8fd

def test_plot_rgb(self, dataset: EuroSAT, tmp_path: Path) -> None:

for example.

JacobJeppesen · 2024-11-30T14:55:58Z

torchgeo/datamodules/flair2.py

+            K.Normalize(mean=self.mean, std=self.std), data_keys=['image', 'mask']
+        )
+
+        self.augs = augs if augs is not None else self.aug


Is self.augs intended to act on the data somewhere? My immediate thought was that it would be related to the augmentations part of the base datamodule:

torchgeo/torchgeo/datamodules/geo.py

Lines 70 to 79 in fedf993

# Data augmentation

Transform = Callable[[dict[str, Tensor]], dict[str, Tensor]]

self.aug: Transform = K.AugmentationSequential(

K.Normalize(mean=self.mean, std=self.std), data_keys=None, keepdim=True

)

self.train_aug: Transform | None = None

self.val_aug: Transform | None = None

self.test_aug: Transform | None = None

self.predict_aug: Transform | None = None

But it's not in there, and setting it to an arbitrary value doesn't seem to do anything. Or perhaps I'm missing something?

JacobJeppesen · 2024-12-01T09:05:25Z

torchgeo/datasets/flair2.py

+    def _load_image(self, path: Path) -> Tensor:
+        """Load a single image.
+
+        Args:
+            path: path to the image
+
+        Returns:
+            Tensor: the loaded image
+        """
+        with rasterio.open(path) as f:
+            array: np.typing.NDArray[np.int_] = f.read()
+            tensor = torch.from_numpy(array).float()
+            if 'B05' in self.bands:
+                # Height channel will always be the last dimension
+                tensor[-1] = torch.div(tensor[-1], 5)
+
+        return tensor


Should the bands perhaps be extracted here based on self.bands? E.g., something like

def _load_image(self, path: Path) -> Tensor: """Load a single image. Args: path: path to the image Returns: Tensor: the loaded image """ with rasterio.open(path) as f: array: np.typing.NDArray[np.int_] = f.read() tensor = torch.from_numpy(array).float() if 'B05' in self.bands: # Height channel will always be the last dimension tensor[-1] = torch.div(tensor[-1], 5) # Extract the bands to be used. E.g., self.bands=("B01", "B02", "B03") will extract the RGB bands. tensor = tensor[[int(band[-2:]) - 1 for band in self.bands]] return tensor

Then when a user has defined n bands in self.bands, the returned sample will only contain those bands, instead of all five. Perhaps self.bands should also be renamed to self.aerial_bands, and a self.sentinel_bands should be added(?)

Mathias Baumgartinger and others added 30 commits September 13, 2024 12:14

docs: update recommended strategy for models with input and output (i…

b96c78b

…mg and msk) Updates in the custom raster dataset tutorial and the actual file documentation. The previous recommended approach (overriding `__get_item__`) is outdated. Refs: microsoft#2292 (reply in thread)

fix: grammar and formatting

0ce9b78

Co-authored-by: Adam J. Stewart <[email protected]>

fix: grammar

d48acd7

Co-authored-by: Adam J. Stewart <[email protected]>

Merge branch 'microsoft:main' into main

8bafb95

feat/WIP: draft for data.py generation file for FLAIR2 dataset

431220d

fix/WIP: fix formatting conventions and populate directories for each…

c9d989d

… type individually

fix/WIP: use alternative hashing algorithm for reproducibility and co…

8c16f69

…rrect rng ranges

feat/WIP: draft for datasets and docs

3f7fec7

Not fully functioning yet, contains copy paste from other datasets

refactor/WIP: mark TODOS, add some documentation

7c4eaf9

feat/WIP: first draft for test FLAIR2

21e5b08

feat: add FLAIR2 import to __init.py__

541630f

feat/WIP: adds sentinel 2 loading/plotting logic

4c1602f

Additionally, some small code refactors are done

Merge branch 'microsoft:main' into main

da45219

feat/WIP: update flair2 unit tests

4bdbe9a

fix&refactor/WIP: provide correct download address for toy dataset an…

8d6d687

…d refine plotting

fix/WIP: properly crop sentinel 2 data

563f7f4

Using the entire sentinel-2 image and a matplotlib patch to debug, otherwise it is really hard to find correct spot due to low resolution

fix/WIP: update cropping slices to match the actual size, fix `_verif…

6aeff90

…y()` for sentinel With the nested dict, it was not possible to download dynamically

feat/WIP: add test data creation of sentinel files

592c336

fix: update tests for FLAIR2 dataset

6d3a189

feat: add dummy data for flair2

9930f6a

fix: properly expose FLAIR2 dataet in __init__.py

ddef891

docs: update documentation of flair2 dataset

63ddc2f

feat: proper integrity checks using correct md5s

ed5ed59

feat: expose an option for using/not using sentinel data

a5af101

feat: add flair2 datamodule

6040cd1

Merge branch 'microsoft:main' into main

2704b3f

feat: new dummy data

96a2207

refactor: syntax for mypy and ruff

6e3301e

fix: bug where sentinel data could be of dimension 0 in first T dimen…

09d5fd6

…sion

feat: save md5s of newly created zips to txt

3393953

md5s might change due to timestamps, this eases the process of changing md5

JacobJeppesen reviewed Nov 25, 2024

View reviewed changes

fix: naming inconsistencies in download url and file name

7ceb0fd

flair_2_centroids_sp_to_patch.zip vs. flair-2_centroids_sp_to_patch.json Refs: microsoft#2394 (comment)

Mathias Baumgartinger added 3 commits November 25, 2024 14:58

fix: format glob-string (i.e. SEN_*{0}.npy) to prevent continously …

ffb7ed0

…extracting Refs: microsoft#2394 (comment)

refactor: different approach to loading sentinels more clearly seperated

9961a5d

Instead of loading both sentinel-data/cloudsnowmasks into one sample, store them sperately. Same with the crop_indices.

refactor: ruff/mypy errors

53b098a

JacobJeppesen reviewed Nov 26, 2024

View reviewed changes

fix: do not assume 1:1 mapping from sentinel to aerial

81dc7d2

This bug caused to omit a good 90% of all images. Refs: microsoft#2394 (comment)

Mathias Baumgartinger added 3 commits November 27, 2024 23:06

docs: remove unwanted BOM from utf-8 encoding in csv file

1c2ca19

Otherwise the entire file is changed, hence this makes the git-history more traceable

feat: expose toy dataset to flair module via flag

86f2d2f

fix: expose FLAIR2Toy dataset under the all section in __init__.py

5673060

Mathias Baumgartinger and others added 3 commits November 28, 2024 15:28

feat: update data generation for 1 to n mapping for sentinel to aeria…

676db57

…l and minor bug fixes

feat: 1:n mapping from sentinel to aerial in tests, also build in nam…

efc4cfd

…ing inconsistencies in original dataset Naming inconsistencies: `flair-2_centroids_sp_to_patch.json` vs `flair_2_centroids_sp_to_patch.zip`

Merge branch 'main' into main

25b1060

MathiasBaumgartinger changed the title ~~[DRAFT] FLAIR#2 Dataset and Datamodule Integration~~ FLAIR#2 Dataset and Datamodule Integration Nov 28, 2024

Mathias Baumgartinger added 2 commits November 28, 2024 18:00

fix: retry removing changes in the entire file

a33c143

Refs: 1c2ca19

refactor: sorting of __all__ tuple in ascending order (ruff)

8779f89

refactor: ruff 0.8 formatting

68a29e6

nilsleh reviewed Nov 29, 2024

View reviewed changes

JacobJeppesen reviewed Nov 30, 2024

View reviewed changes

JacobJeppesen reviewed Dec 1, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

FLAIR#2 Dataset and Datamodule Integration #2394

FLAIR#2 Dataset and Datamodule Integration #2394

MathiasBaumgartinger commented Nov 5, 2024 •

edited

Loading

JacobJeppesen Nov 25, 2024

JacobJeppesen Nov 25, 2024

MathiasBaumgartinger Nov 25, 2024

JacobJeppesen Nov 25, 2024

JacobJeppesen Nov 26, 2024

JacobJeppesen commented Nov 25, 2024

MathiasBaumgartinger commented Nov 25, 2024

JacobJeppesen commented Nov 25, 2024 •

edited

Loading

JacobJeppesen Nov 26, 2024

MathiasBaumgartinger Nov 26, 2024

JacobJeppesen Nov 26, 2024

JacobJeppesen commented Nov 26, 2024

adamjstewart commented Nov 26, 2024

JacobJeppesen commented Nov 26, 2024

JacobJeppesen commented Nov 28, 2024

adamjstewart commented Nov 28, 2024

MathiasBaumgartinger commented Nov 28, 2024

adamjstewart commented Nov 28, 2024

nilsleh Nov 29, 2024

nilsleh Nov 29, 2024

nilsleh Nov 29, 2024

nilsleh Nov 29, 2024

MathiasBaumgartinger Nov 29, 2024

nilsleh Nov 29, 2024

MathiasBaumgartinger Nov 29, 2024

nilsleh Nov 29, 2024

nilsleh Nov 29, 2024

nilsleh Nov 29, 2024

nilsleh Nov 29, 2024

JacobJeppesen Nov 30, 2024

JacobJeppesen Dec 1, 2024

		@@ -0,0 +1,79 @@
		"""This module contains the FLAIR2DataModule class for loading the FLAIR2 dataset.

		@@ -0,0 +1,8 @@
		/home/mathias/Dev/forks/torchgeo/tests/data/flair2/FLAIR2/flair_2_labels_test.zip: b13c4a3cb7ebb5cadddc36474bb386f9

	# Data augmentation
	Transform = Callable[[dict[str, Tensor]], dict[str, Tensor]]
	self.aug: Transform = K.AugmentationSequential(
	K.Normalize(mean=self.mean, std=self.std), data_keys=None, keepdim=True
	)

	self.train_aug: Transform \| None = None
	self.val_aug: Transform \| None = None
	self.test_aug: Transform \| None = None
	self.predict_aug: Transform \| None = None

FLAIR#2 Dataset and Datamodule Integration #2394

Are you sure you want to change the base?

FLAIR#2 Dataset and Datamodule Integration #2394

Conversation

MathiasBaumgartinger commented Nov 5, 2024 • edited Loading

FLAIR#2 dataset

Implementation Details

NonGeoDataset, __init()__

_verify, _download, _extract

_load_image, _load_sentinel, _load_target

Questions

TODOs/FIXMEs

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

JacobJeppesen commented Nov 25, 2024

MathiasBaumgartinger commented Nov 25, 2024

JacobJeppesen commented Nov 25, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

JacobJeppesen commented Nov 26, 2024

adamjstewart commented Nov 26, 2024

JacobJeppesen commented Nov 26, 2024

JacobJeppesen commented Nov 28, 2024

adamjstewart commented Nov 28, 2024

MathiasBaumgartinger commented Nov 28, 2024

adamjstewart commented Nov 28, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

MathiasBaumgartinger commented Nov 5, 2024 •

edited

Loading

`NonGeoDataset`, `init()`

`_verify`, `_download`, `_extract`

`_load_image`, `_load_sentinel`, `_load_target`

JacobJeppesen commented Nov 25, 2024 •

edited

Loading