Add CHC dataset

DATASET_GUIDE.md

@@ -24,6 +24,7 @@ This document provides a detailed overview of the datasets used in this reposito
   - [Multi-label Classification (e.g., m-BigEarthNet)](#for-multi-label-classification-eg-m-bigearthnet)
   - [Single-label Classification (e.g., m-EuroSat, m-Brick-Kiln)](#for-single-label-classification-ie-m-eurosat-m-brick-kiln-m-forestnet-m-pv4ger-m-so2sat)
   - [Semantic Segmentation (e.g., m-NZ-Cattle, m-SA-Crop-Type)](#for-semantic-segmentation-ie-m-cashew-plantation-m-chesapeake-landcover-m-neontree-m-nz-cattle-m-pv4ger-seg-and-m-sa-crop-type)
+- [CHC](#chc)
 
 ---
 
@@ -350,4 +351,16 @@ Note that `export GEO_BENCH_DIR=YOUR/PATH/DIR` is required.
     - The criterion is set to `none` since no training is performed
     - The batch size can be larger since we're only doing inference
     - `finetune` is set to `false` as we're only using the pre-trained encoder
-    - The task is set to `knn_probe_multi_label` to handle multiple labels per sample
+    - The task is set to `knn_probe_multi_label` to handle multiple labels per sample
+
+### CHC
+   ```
+    torchrun --nnodes=1 --nproc_per_node=1 pangaea/run.py \
+    --config-name=train \
+    dataset=chc \
+    encoder=dofa \
+    decoder=reg_upernet_mt_ltae \
+    preprocessing=reg_default \
+    criterion=mse \
+    task=regression
+  ```

configs/dataset/chc.yaml

@@ -0,0 +1,39 @@
+_target_: pangaea.datasets.chc.CHCDataset
+dataset_name: CHCDataset
+root_path: ./data/chc
+download_url: https://sid.erda.dk/share_redirect/H4n8mswu1F
+auto_download: True
+img_size: 1024
+multi_temporal: 6
+multi_modal: False
+ignore_index: -1
+num_classes: 1
+classes:
+  - regression
+
+distribution:
+  - 1.
+
+bands:
+  optical:
+    - B2 # B
+    - B3 # G
+    - B4 # R
+    - B8A # NIR
+data_mean:
+  optical: [514.16294, 849.4592, 1033.3091, 2873.6320]
+data_std:
+  optical: [316.0711, 376.3348, 568.9334, 758.5220]
+data_min:
+  optical: [1., 1., 1., 1.]
+data_max:
+  optical: [15535., 14703., 28714., 59879.]
+
+img_merging: mean
+min_height: 3.
+confidence_z_value: 1.6  # corresponds to 90% confidence interval
+oversampling_factor: 1
+rm_vegetation_edges: True
+masking:
+  - veg
+weighting: none

environment.yaml

@@ -31,4 +31,4 @@ dependencies:
   - yacs
   - wandb
   - hydra-core
-
+  - scikit-image