Isaac Corley

Replace open3d with pyvista

1

Refactors point cloud plotting in the `torchgeo.datasets.IDTReeS` dataset to use `pyvista` instead of `open3d` Note: Currently figuring out the min version which passes tests. See #662 Example: https://user-images.githubusercontent.com/22203655/178155777-dac2e462-d978-4005-9412-3ffd91a10bcb.mov

Urban 3D Challenge Dataset

6

This PR adds the [Urban 3D Challenge dataset](https://spacenet.ai/the-ussocom-urban-3d-competition/) from the USSOCOM Urban 3D Competition (2017). This dataset combines 2D RGB imagery with 3D Digital Terrain Models (DTM) and Digital Surface...

This PR adds the [PASTIS dataset](https://github.com/VSainteuf/pastis-benchmark) (original and PASTIS-R versions) for time-series semantic/instance segmentation of agricultural parcels of 18 crop type categories in Sentinel-1 and Sentinel-2 imagery.  TODO: -...

Add Transforms Benchmark

6

- Add `benchmark_transforms.py` which allows one to compare the execution time of a sequential pipeline of our transforms + kornia augmentations for CPU vs GPU and various batch image/mask sizes.

[FEATURE] Vision Transformer Feature Extraction

7

It seems that ```VisionTransformer``` doesn't support feature extraction of all outputs in the ```forward_features``` method. Only returning of the cls token or [cls_token, distillation_token] is available [timm/models/vision_transformer.py#L291-L304](https://github.com/rwightman/pytorch-image-models/blob/23c18a33e4168dc7cb11439c1f9acd38dc8e9824/timm/models/vision_transformer.py#L291-L304). This functionality seems...

The `DecoderBlock` used in `UnetDecoder` hardcodes the upsampling factor to 2. This works for ResNet encoders however this is problematic for models like ConvNext which downsample by a factor of...

NLCD2016 Tree Canopy

8

This PR adds the NLCD2016 Tree Canopy dataset See https://www.mrlc.gov/data/nlcd-2016-usfs-tree-canopy-cover-conus

This PR adds the Cross-City Multimodal Semantic Segmentation Challenge (C2Seg) dataset for the WHISPERS2023 conference. The dataset contains a mix of Sentinel-1, Sentinel-2, EnMAP HSI, and Gaofen MSI/SAR/HSI imagery. [Link...

fix SeCo transforms

11

The current SeCo transforms do the following: - resize - center crop - normalize using the quantiles - multiply by 255 - normalize using imagenet stats However in the [script](https://github.com/ServiceNow/seasonal-contrast/blob/8285173ec205b64bc3e53b880344dd6c3f79fa7a/datasets/seco_dataset.py)...