Merge pull request #400 from nasaharvest/Add-data-Kenya-Crop-Area-2019

Add Kenya Crop Area Data

data/datasets.dvc

@@ -1,6 +1,6 @@
 outs:
-- md5: add693a42b38c020b895e8e7c855210d.dir
-  size: 709637820
-  nfiles: 55
+- md5: a250ce1c98882904e609f7abc88ed404.dir
+  size: 712601175
+  nfiles: 56
   path: datasets
   hash: md5

data/raw.dvc

@@ -1,6 +1,6 @@
 outs:
-- md5: 7b5c138e97f360400e6b3fc080d896e2.dir
-  size: 445019716
-  nfiles: 394
+- md5: bb6ed42f700db44dba28965972d936cf.dir
+  size: 445220011
+  nfiles: 396
   path: raw
   hash: md5

data/report.txt

@@ -457,3 +457,12 @@ UgandaNorthCorLabel2022 (Timesteps: 24)
 ----------------------------------------------------------------------------
 eo_data_complete    2975
 ✔ training amount: 2975, positive class: 51.6%
+
+
+
+KenyaCropArea2019 (Timesteps: 24)
+----------------------------------------------------------------------------
+eo_data_complete    544
+✔ training amount: 232, positive class: 23.7%
+✔ validation amount: 149, positive class: 18.1%
+✔ testing amount: 163, positive class: 20.2%

datasets.py

@@ -552,6 +552,38 @@ def load_labels(self) -> pd.DataFrame:
         return df
 
 
+class KenyaCropArea2019(LabeledDataset):
+    def load_labels(self) -> pd.DataFrame:
+        raw_folder = raw_dir / "Kenya_Crop_Area_2019"
+        df1 = pd.read_csv(
+            raw_folder / "ceo-Kenya-Crop-Area-Estimation-Reference-Sample-2019---Set-1"
+            "-sample-data-2024-06-10.csv"
+        )
+        df2 = pd.read_csv(
+            raw_folder / "ceo-Kenya-Crop-Area-Estimation-Reference-Sample-2019---Set-2"
+            "-sample-data-2024-06-10.csv"
+        )
+        df = pd.concat([df1, df2])
+
+        # Discard rows with no label
+        df = df[~df["Does this point correspond to active cropland?"].isna()].copy()
+        df[CLASS_PROB] = df["Does this point correspond to active cropland?"] == "Crop"
+        df[CLASS_PROB] = df[CLASS_PROB].astype(int)
+        df["num_labelers"] = 1
+        df = df.groupby([LON, LAT], as_index=False, sort=False).agg(
+            {
+                CLASS_PROB: "mean",
+                "num_labelers": "sum",
+                "plotid": join_unique,
+                "sampleid": join_unique,
+                "email": join_unique,
+            }
+        )
+        df[START], df[END] = date(2019, 1, 1), date(2020, 12, 31)
+        df[SUBSET] = train_val_test_split(df.index, 0.3, 0.3)
+        return df
+
+
 datasets: List[LabeledDataset] = [
     CustomLabeledDataset(
         dataset="geowiki_landcover_2017",
@@ -1340,6 +1372,7 @@ def load_labels(self) -> pd.DataFrame:
     Uganda_NorthCEO2021(),
     UgandaNorthCEO2019(),
     UgandaNorthCorLabel2022(),
+    KenyaCropArea2019(),
 ]
 
 if __name__ == "__main__":