merge master branch into this branch

nasaharvest · Feb 26, 2024 · f6e1600 · f6e1600
2 parents 52933cf + 120282b
commit f6e1600
Show file tree

Hide file tree

Showing 10 changed files with 2,550 additions and 129 deletions.
diff --git a/data/all_dataset_params.json b/data/all_dataset_params.json
@@ -808,6 +808,56 @@
             12
         ]
     },
+    "geowiki_landcover_2017,Kenya,Mali,Mali_lower_CEO_2019,Mali_upper_CEO_2019,Togo,Rwanda,Uganda,open_buildings,digitalearthafrica_eastern,digitalearthafrica_sahel,Ethiopia,Ethiopia_Tigray_2020,Ethiopia_Tigray_2021,Ethiopia_Bure_Jimma_2019,Ethiopia_Bure_Jimma_2020,Malawi_CEO_2020,Malawi_CEO_2019,Malawi_FAO,Malawi_FAO_corrected,Zambia_CEO_2019,Tanzania_CEO_2019,Namibia_corrective_labels_2020,Malawi_corrected,Namibia_CEO_2020,Namibia_WFP,Sudan_Blue_Nile_CEO_2019,Hawaii_CEO_2020,Senegal_CEO_2022,HawaiiAgriculturalLandUse2020,KenyaCEO2019,HawaiiCorrective2020,HawaiiCorrectiveGuided2020,MalawiCorrectiveLabels2020,SudanBlueNileCEO2020,SudanBlueNileCorrectiveLabels2019,EthiopiaTigrayCorrective2020,SudanAlGadarefCEO2019,MaliStratifiedCEO2019,SudanAlGadarefCEO2020,NamibiaNorthStratified2020,Namibia_field_samples_22_23,SudanGedarefDarfurAlJazirah2022,Uganda_NorthCEO2022,Uganda_NorthCEO2021_February_2022": {
+        "normalizing_dict": {
+            "mean": [
+                -11.253425188439937,
+                -18.32938103852523,
+                1435.394781661943,
+                1401.2376604488686,
+                1424.1887305810615,
+                1659.930606706183,
+                2418.9610242192266,
+                2797.4266694196153,
+                2686.556230689381,
+                3059.6146535391536,
+                839.3203256303339,
+                2492.394136925002,
+                1656.5102430595289,
+                290.4116372888038,
+                0.004740448735646688,
+                862.6486426458467,
+                5.645431586296536,
+                0.3394156161911947
+            ],
+            "std": [
+                4.148125981435058,
+                5.060516119375943,
+                994.0712142249523,
+                965.2564704509458,
+                1160.026751746973,
+                1108.4164346403447,
+                1050.1378976686253,
+                1116.0387648387118,
+                1073.380355703031,
+                1145.5432935794531,
+                654.3036779889275,
+                1165.4722835564185,
+                1054.8715358765717,
+                35.62925408430905,
+                0.014667963835620806,
+                646.2067949711964,
+                7.102936150840518,
+                0.22947527099163295
+            ]
+        },
+        "train_num_timesteps": [
+            12
+        ],
+        "val_num_timesteps": [
+            12
+        ]
+    },
     "geowiki_landcover_2017,Kenya,Mali,Mali_lower_CEO_2019,Mali_upper_CEO_2019,Togo,Rwanda,Uganda,open_buildings,digitalearthafrica_eastern,digitalearthafrica_sahel,Ethiopia,Ethiopia_Tigray_2020,Ethiopia_Tigray_2021,Ethiopia_Bure_Jimma_2019,Ethiopia_Bure_Jimma_2020,Malawi_CEO_2020,Malawi_CEO_2019,Malawi_FAO,Malawi_FAO_corrected,Zambia_CEO_2019,Tanzania_CEO_2019,Namibia_corrective_labels_2020,Malawi_corrected,Namibia_CEO_2020,Namibia_WFP,Sudan_Blue_Nile_CEO_2019,Hawaii_CEO_2020,Senegal_CEO_2022,HawaiiAgriculturalLandUse2020,KenyaCEO2019,HawaiiCorrective2020,HawaiiCorrectiveGuided2020,MalawiCorrectiveLabels2020,SudanBlueNileCEO2020,SudanBlueNileCorrectiveLabels2019,EthiopiaTigrayCorrective2020,SudanAlGadarefCEO2019,MaliStratifiedCEO2019,SudanAlGadarefCEO2020,NamibiaNorthStratified2020,Namibia_field_samples_22_23,SudanGedarefDarfurAlJazirah2022,Uganda_NorthCEO2022_February_2022": {
         "normalizing_dict": {
             "mean": [

diff --git a/data/datasets.dvc b/data/datasets.dvc
@@ -1,6 +1,6 @@
 outs:
-- md5: 63c6f22459e60941aa459671fbff891b.dir
-  size: 671350235
-  nfiles: 48
+- md5: b7c381babff36cca1ef539c77e677de4.dir
+  size: 676810981
+  nfiles: 49
   path: datasets
   hash: md5
diff --git a/data/models.dvc b/data/models.dvc
@@ -1,6 +1,6 @@
 outs:
-- md5: 3c2840cad82955c5d02cbc3ef60757a0.dir
-  size: 82333214
-  nfiles: 66
+- md5: e2f4f187db0f865868e57c4b69d1409c.dir
+  size: 83854781
+  nfiles: 68
   path: models
   hash: md5
diff --git a/data/models.json b/data/models.json
@@ -356,6 +356,23 @@
             "roc_auc_score": 0.8442
         }
     },
+    "Uganda_North_2021_V1": {
+        "params": "https://wandb.ai/nasa-harvest/crop-mask/runs/jecuc6f8",
+        "test_metrics": {
+            "accuracy": 0.8308,
+            "f1_score": 0.5545,
+            "precision_score": 0.6364,
+            "recall_score": 0.4912,
+            "roc_auc_score": 0.8323
+        },
+        "val_metrics": {
+            "accuracy": 0.8355,
+            "f1_score": 0.4865,
+            "precision_score": 0.5625,
+            "recall_score": 0.4286,
+            "roc_auc_score": 0.8153
+        }
+    },
     "Uganda_North_2022_V1": {
         "params": "https://wandb.ai/nasa-harvest/crop-mask/runs/v5f8m489",
         "test_metrics": {

diff --git a/data/raw.dvc b/data/raw.dvc
@@ -1,6 +1,6 @@
 outs:
-- md5: c033b32e0901a197d9f743c8329b24ae.dir
-  size: 443888649
-  nfiles: 387
+- md5: 09700b46df6451c928d36156bd01507a.dir
+  size: 444212594
+  nfiles: 389
   path: raw
   hash: md5
diff --git a/data/report.txt b/data/report.txt
@@ -422,3 +422,12 @@ eo_data_complete    1000
 ✔ training amount: 387, positive class: 20.9%
 ✔ validation amount: 294, positive class: 23.1%
 ✔ testing amount: 319, positive class: 17.6%
+
+
+
+Uganda_NorthCEO2021 (Timesteps: 24)
+----------------------------------------------------------------------------
+eo_data_complete    1000
+✔ training amount: 387, positive class: 16.0%
+✔ validation amount: 294, positive class: 14.3%
+✔ testing amount: 319, positive class: 17.9%
diff --git a/datasets.py b/datasets.py
@@ -448,6 +448,38 @@ def load_labels(self) -> pd.DataFrame:
         return df
 
 
+class Uganda_NorthCEO2021(LabeledDataset):
+    def load_labels(self) -> pd.DataFrame:
+        raw_folder = raw_dir / "Uganda_North_2021"
+        df1 = pd.read_csv(
+            raw_folder
+            / "ceo-UNHCR-North-Uganda-Feb-2021---Feb-2022-(Set-1)-sample-data-2024-02-07.csv"
+        )
+        df2 = pd.read_csv(
+            raw_folder
+            / "ceo-UNHCR-North-Uganda-Feb-2021---Feb-2022-(Set-2)-sample-data-2024-02-07.csv"
+        )
+        df = pd.concat([df1, df2])
+
+        # Discard rows with no label
+        df = df[~df["Does this pixel contain active cropland?"].isna()].copy()
+        df[CLASS_PROB] = df["Does this pixel contain 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(2021, 1, 1), date(2022, 12, 31)
+        df[SUBSET] = train_val_test_split(df.index, 0.3, 0.3)
+        return df
+
+
 datasets: List[LabeledDataset] = [
     CustomLabeledDataset(
         dataset="geowiki_landcover_2017",
@@ -1232,6 +1264,7 @@ def load_labels(self) -> pd.DataFrame:
     Namibia_field_samples_22_23(),
     SudanGedarefDarfurAlJazirah2022(),
     Uganda_NorthCEO2022(),
+    Uganda_NorthCEO2021(),
 ]
 
 if __name__ == "__main__":