feat: implement Spatio-Temporal Deep Learning pipeline for inference worker

- Add hybrid PyTorch (TemporalFCN) + CatBoost ensemble logic in new hybrid_inference.py
- Update worker.py to support 'Hybrid' model type with artifact syncing from MinIO
- Integrate odc-stac for raw spectral index fetching from DE Africa STAC
- Update requirements.txt with torch, odc-stac, and rioxarray dependencies
- Include ntfy deployment in k8s manifests

apps/worker/hybrid_inference.py

@@ -0,0 +1,276 @@
+import os
+import io
+import json
+import time
+import copy
+import joblib
+import torch
+import torch.nn as nn
+from torch.utils.data import Dataset, DataLoader
+import pandas as pd
+import numpy as np
+from sklearn.neighbors import KNeighborsRegressor
+from catboost import CatBoostClassifier
+
+# Digital Earth Africa STAC specific imports
+try:
+    from pystac_client import Client
+    import odc.stac
+    import xarray as xr
+    import rioxarray
+except ImportError:
+    Client = None
+    odc = None
+    xr = None
+    rioxarray = None
+
+# ==========================================
+# 1. CPU-OPTIMIZED ARCHITECTURES
+# ==========================================
+
+class TemporalFCN(nn.Module):
+    def __init__(self, num_bands, num_classes):
+        super().__init__()
+        self.conv_block1 = nn.Sequential(
+            nn.Conv1d(num_bands, 64, kernel_size=5, padding=2),
+            nn.BatchNorm1d(64),
+            nn.ReLU()
+        )
+        self.conv_block2 = nn.Sequential(
+            nn.Conv1d(64, 128, kernel_size=3, padding=1),
+            nn.BatchNorm1d(128),
+            nn.ReLU()
+        )
+        self.global_avg_pool = nn.AdaptiveAvgPool1d(1)
+        self.fc = nn.Linear(128, num_classes)
+
+    def forward(self, x, return_features=False):
+        x = self.conv_block1(x)
+        x = self.conv_block2(x)
+        features = self.global_avg_pool(x).squeeze(-1)
+        out = self.fc(features)
+        if return_features:
+            return out, features
+        return out
+
+class SmallGRU(nn.Module):
+    def __init__(self, num_bands, num_classes, hidden_size=64):
+        super().__init__()
+        self.gru = nn.GRU(input_size=num_bands, hidden_size=hidden_size, num_layers=1, batch_first=True)
+        self.fc = nn.Linear(hidden_size, num_classes)
+
+    def forward(self, x, return_features=False):
+        x = x.transpose(1, 2)
+        out, _ = self.gru(x)
+        features = out[:, -1, :]
+        final_out = self.fc(features)
+        if return_features:
+            return final_out, features
+        return final_out
+
+# ==========================================
+# 2. DATA PREPARATION & PYTORCH UTILS
+# ==========================================
+
+class CropDataset(Dataset):
+    def __init__(self, X, y, augment=False):
+        self.X = torch.FloatTensor(X)
+        self.y = torch.LongTensor(y)
+        self.augment = augment
+
+    def __len__(self):
+        return len(self.y)
+
+    def __getitem__(self, idx):
+        x = self.X[idx].clone()
+        if self.augment:
+            if torch.rand(1).item() > 0.5:
+                noise = torch.randn_like(x) * 0.03
+                x = x + noise
+            if torch.rand(1).item() > 0.7:
+                seq_len = x.shape[1]
+                t_idx = torch.randint(0, seq_len, (1,)).item()
+                x[:, t_idx] = 0.0
+        return x, self.y[idx]
+
+def prepare_tensors(df, bands, dates):
+    num_samples = len(df)
+    X_3d = np.zeros((num_samples, len(bands), len(dates)), dtype=np.float32)
+    for b_idx, band in enumerate(bands):
+        for d_idx, date in enumerate(dates):
+            col = f"{date}_{band}"
+            if col in df.columns:
+                X_3d[:, b_idx, d_idx] = df[col].values
+    
+    means = X_3d.mean(axis=2, keepdims=True)
+    stds = X_3d.std(axis=2, keepdims=True) + 1e-8
+    X_3d = (X_3d - means) / stds
+    return X_3d
+
+# ==========================================
+# 3. DIGITAL EARTH AFRICA STAC INTEGRATION
+# ==========================================
+
+class DEAfricaSTACWrapper:
+    def __init__(self, stac_url="https://explorer.digitalearth.africa/stac"):
+        if Client is None or odc is None or xr is None:
+            raise ImportError("Missing required libraries: pystac-client, odc-stac, xarray")
+        print(f"Connecting to Digital Earth Africa STAC Catalog at {stac_url}...")
+        self.catalog = Client.open(stac_url)
+
+    def fetch_and_format_data(self, lat_range, lon_range, time_range, resolution=20):
+        bbox = [lon_range[0], lat_range[0], lon_range[1], lat_range[1]]
+        print(f"Searching STAC for Bounding Box: {bbox} over {time_range}...")
+        search = self.catalog.search(
+            collections=["s2_l2a"],
+            bbox=bbox,
+            datetime=f"{time_range[0]}/{time_range[1]}"
+        )
+        items = list(search.items())
+        if not items:
+            raise ValueError("No STAC items found for this bounding box and time range.")
+        
+        print(f"Found {len(items)} STAC items. Loading into xarray...")
+        ds = odc.stac.load(
+            items,
+            measurements=['red', 'green', 'blue', 'nir', 'red_edge_1', 'scl'],
+            bbox=bbox,
+            crs="EPSG:6933",
+            resolution=resolution,
+            groupby="solar_day"
+        )
+        
+        print("Masking clouds and shadows...")
+        valid_mask = (ds.scl == 4) | (ds.scl == 5) | (ds.scl == 6) | (ds.scl == 2) | (ds.scl == 7)
+        ds = ds.where(valid_mask)
+        ds = ds / 10000.0
+        
+        print("Computing Spectral Indices (NDVI, NDRE, SAVI, EVI)...")
+        ds['ndvi'] = (ds.nir - ds.red) / (ds.nir + ds.red + 1e-8)
+        ds['ndre'] = (ds.nir - ds.red_edge_1) / (ds.nir + ds.red_edge_1 + 1e-8)
+        ds['savi'] = ((ds.nir - ds.red) / (ds.nir + ds.red + 0.5)) * 1.5
+        ds['evi'] = 2.5 * ((ds.nir - ds.red) / (ds.nir + 6 * ds.red - 7.5 * ds.blue + 1))
+        
+        ds_indices = ds[['ndvi', 'ndre', 'evi', 'savi']]
+        
+        print("Reshaping multi-dimensional xarray into flat Tabular DataFrame...")
+        df = ds_indices.compute().to_dataframe().reset_index()
+        df['date_str'] = df['time'].dt.strftime('%Y%m%d')
+        
+        df_pivot = df.pivot(index=['y', 'x'], columns='date_str', values=['ndvi', 'ndre', 'evi', 'savi'])
+        df_pivot.columns = [f"{date}_{band}" for band, date in df_pivot.columns]
+        
+        df_final = df_pivot.reset_index().rename(columns={'y': 'lat', 'x': 'lon'})
+        print(f"✅ Data Ready! {df_final.shape[0]} spatial pixels generated.")
+        return df_final
+
+# ==========================================
+# 4. INFERENCE PIPELINE
+# ==========================================
+
+class CropInferencePipeline:
+    def __init__(self, model_dir="/tmp/geocrop-cache"):
+        print(f"Loading Crop Inference Pipeline from {model_dir}...")
+        meta_path = os.path.join(model_dir, "pipeline_meta.pkl")
+        if not os.path.exists(meta_path):
+            raise FileNotFoundError(f"Pipeline metadata not found at {meta_path}")
+            
+        self.meta = joblib.load(meta_path)
+        self.le = self.meta["le"]
+        self.bands = self.meta["bands"]
+        self.dates = self.meta["dates"]
+        self.w_fcn = self.meta["weights"]["w_fcn"]
+        self.w_cb = self.meta["weights"]["w_cb"]
+        
+        self.fcn = TemporalFCN(len(self.bands), self.meta["num_classes"])
+        fcn_path = os.path.join(model_dir, "Temporal_FCN.pth")
+        self.fcn.load_state_dict(torch.load(fcn_path, map_location=torch.device('cpu')))
+        self.fcn.eval()
+        
+        cb_path = os.path.join(model_dir, "calibrated_hybrid_cb.pkl")
+        self.calibrated_cb = joblib.load(cb_path)
+        print("Models loaded successfully.")
+
+    def _impute_inference_data(self, df):
+        print("Imputing cloudy/missing timesteps via temporal interpolation...")
+        df = df.copy()
+        missing_mask = {}
+        for band in self.bands:
+            band_cols = [f"{date}_{band}" for date in self.dates if f"{date}_{band}" in df.columns]
+            if band_cols:
+                missing_mask[band] = df[band_cols].isna().astype(float)
+                df[band_cols] = df[band_cols].interpolate(method='linear', axis=1, limit_direction='both')
+                df[band_cols] = df[band_cols].ffill(axis=1).bfill(axis=1).fillna(0)
+        return df, missing_mask
+
+    def predict(self, raw_df, apply_spatial_smoothing=False, coord_cols=['lat', 'lon']):
+        df, missing_mask = self._impute_inference_data(raw_df)
+        X_infer = prepare_tensors(df, self.bands, self.dates)
+        infer_loader = DataLoader(CropDataset(X_infer, np.zeros(len(df)), augment=False), batch_size=128, shuffle=False)
+        
+        fcn_probs = []
+        fcn_feats = []
+        with torch.no_grad():
+            for X_batch, _ in infer_loader:
+                out, feats = self.fcn(X_batch, return_features=True)
+                fcn_probs.extend(torch.softmax(out, dim=1).numpy())
+                fcn_feats.append(feats.numpy())
+        
+        fcn_probs = np.array(fcn_probs)
+        fcn_feats = np.vstack(fcn_feats)
+        
+        X_infer_flat = X_infer.reshape(X_infer.shape[0], -1)
+        X_stack = np.hstack([X_infer_flat, fcn_feats])
+        cb_probs = self.calibrated_cb.predict_proba(X_stack)
+        
+        final_probs = (fcn_probs * self.w_fcn) + (cb_probs * self.w_cb)
+        final_preds = np.argmax(final_probs, axis=1)
+        
+        if apply_spatial_smoothing and all(col in df.columns for col in coord_cols):
+            print(f"Applying spatial probability smoothing using {coord_cols}...")
+            coords = df[coord_cols].values
+            knn = KNeighborsRegressor(n_neighbors=9, weights='distance')
+            knn.fit(coords, final_probs)
+            smoothed_probs = knn.predict(coords)
+            final_preds = np.argmax(smoothed_probs, axis=1)
+            final_probs = smoothed_probs
+            
+        df['class_id'] = final_preds
+        df['predicted_crop'] = self.le.inverse_transform(final_preds)
+        df['confidence'] = np.max(final_probs, axis=1)
+        
+        missing_ratio = np.mean([m.mean(axis=1) for m in missing_mask.values()], axis=0)
+        df['high_missing'] = missing_ratio > 0.4
+        df['low_quality'] = (df['confidence'] < 0.5) | df['high_missing']
+        
+        # Set NoData (0) for low quality
+        df.loc[df['low_quality'], 'class_id'] = 0
+        df.loc[df['low_quality'], 'predicted_crop'] = 'Unknown/NoData'
+        return df
+
+    def export_to_geotiff(self, df, output_path="lulc_map.tif", crs="EPSG:6933"):
+        if xr is None or rioxarray is None:
+            raise ImportError("Missing required libraries: xarray, rioxarray")
+        print(f"Exporting LULC masks to {output_path}...")
+        
+        ds_out = df.set_index(['lat', 'lon'])[['class_id', 'confidence', 'low_quality']].to_xarray()
+        ds_out = ds_out.rename({'lat': 'y', 'lon': 'x'})
+        ds_out = ds_out.sortby('y', ascending=False)
+        ds_out = ds_out.rio.set_spatial_dims(x_dim='x', y_dim='y')
+        ds_out.rio.write_crs(crs, inplace=True)
+        
+        ds_out['class_id'].astype('uint16').rio.to_raster(output_path)
+        
+        conf_path = output_path.replace('.tif', '_confidence.tif')
+        ds_out['confidence'].astype('float32').rio.to_raster(conf_path)
+        
+        mask_path = output_path.replace('.tif', '_cloud_mask.tif')
+        ds_out['low_quality'].astype('uint8').rio.to_raster(mask_path)
+        
+        legend_path = output_path.replace('.tif', '_legend.json')
+        legend_dict = {int(i): str(c) for i, c in enumerate(self.le.classes_)}
+        if 0 not in legend_dict:
+            legend_dict[0] = 'Unknown/NoData'
+        with open(legend_path, 'w') as f:
+            json.dump(legend_dict, f, indent=4)
+        print(f"✅ Successfully exported GeoTIFFs and class legend!")

apps/worker/requirements.txt

@@ -5,6 +5,7 @@ rq
 # Core dependencies
 numpy>=1.24.0
 pandas>=2.0.0
+torch --index-url https://download.pytorch.org/whl/cpu
 
 # Raster/geo processing
 rasterio>=1.3.0
@@ -13,6 +14,7 @@ rioxarray>=0.14.0
 # STAC data access
 pystac-client>=0.7.0
 stackstac>=0.4.0
+odc-stac>=0.3.0
 xarray>=2023.1.0
 
 # ML

apps/worker/worker.py

@@ -171,7 +171,7 @@ def parse_and_validate_payload(payload: dict) -> tuple[dict, List[str]]:
     
     # Validate model
     if "model" in payload:
-        valid_models = ["Ensemble", "RandomForest", "XGBoost", "LightGBM", "CatBoost"]
+        valid_models = ["Ensemble", "RandomForest", "XGBoost", "LightGBM", "CatBoost", "CatBoost_V2"]
         if payload["model"] not in valid_models:
             errors.append(f"Invalid model: {payload['model']}. Must be one of {valid_models}")
     
@@ -372,112 +372,82 @@ def run_job(payload_dict: dict) -> dict:
             print(f"[{job_id}] Synthetic feature cube shape: {feature_cube.shape}")
         
         # ==========================================
-        # Stage 3: Load DW Baseline
+        # Stage 3: Load Model Artifacts
         # ==========================================
-        update_status(job_id, "running", "load_dw", 40, "Loading DW baseline...")
+        update_status(job_id, "running", "load_model", 40, "Loading model artifacts...")
         
-        print(f"[{job_id}] Loading DW baseline for {payload['year']}...")
+        is_hybrid = "hybrid" in payload['model'].lower() or "spatiotemporal" in payload['model'].lower()
         
-        from dw_baseline import load_dw_baseline_window
+        model_dir = Path(tempfile.mkdtemp())
         
+        if is_hybrid:
+            print(f"[{job_id}] Model type: Hybrid Spatio-Temporal. Downloading artifacts...")
+            # Expected files in MinIO: pipeline_meta.pkl, Temporal_FCN.pth, calibrated_hybrid_cb.pkl
+            for artifact in ["pipeline_meta.pkl", "Temporal_FCN.pth", "calibrated_hybrid_cb.pkl"]:
                 try:
-            dw_arr, dw_profile = load_dw_baseline_window(
+                    storage.download_model_file(artifact, model_dir)
-                storage=storage,
+                    print(f"[{job_id}] Downloaded {artifact}")
-                year=payload['year'],
-                aoi_bbox_wgs84=bbox,
-                season=payload['season'],
-            )
-            
-            if dw_arr is None:
-                raise FileNotFoundError(f"No DW baseline found for year {payload['year']}")
-            
-            print(f"[{job_id}] DW baseline shape: {dw_arr.shape}")
-            
                 except Exception as e:
-            update_status(
+                    print(f"[{job_id}] Failed to download {artifact}: {e}")
-                job_id, "failed", "load_dw", 45,
+                    # Try with 'hybrid/' prefix if direct fails
-                f"Failed to load DW baseline: {e}",
+                    try:
-                error={"type": "DWBASELINE_ERROR", "message": str(e)}
+                        storage.download_file("geocrop-models", f"hybrid/{artifact}", model_dir / artifact)
+                        print(f"[{job_id}] Downloaded {artifact} (from hybrid/ prefix)")
+                    except Exception as e2:
+                        raise FileNotFoundError(f"Required artifact {artifact} not found in geocrop-models: {e2}")
+            
+            # ==========================================
+            # Stage 4: Fetch Spatio-Temporal Data
+            # ==========================================
+            update_status(job_id, "running", "fetch_stac", 50, "Fetching spatio-temporal indices...")
+            from hybrid_inference import DEAfricaSTACWrapper, CropInferencePipeline
+            
+            stac_wrapper = DEAfricaSTACWrapper()
+            # Calculate ranges for wrapper
+            lat_range = (bbox[1], bbox[3])
+            lon_range = (bbox[0], bbox[2])
+            time_range = (start_date, end_date)
+            
+            unseen_pixel_df = stac_wrapper.fetch_and_format_data(
+                lat_range=lat_range,
+                lon_range=lon_range,
+                time_range=time_range
             )
-            return {"status": "failed", "error": f"DW baseline error: {e}"}
             
             # ==========================================
-        # Stage 4: Skip AI Inference, use DW as result
+            # Stage 5: Hybrid Inference
             # ==========================================
-        update_status(job_id, "running", "infer", 60, "Using DW baseline as classification...")
+            update_status(job_id, "running", "infer", 70, "Running Hybrid Inference (CNN + CatBoost)...")
+            pipeline = CropInferencePipeline(model_dir=str(model_dir))
             
-        print(f"[{job_id}] Using DW baseline as result (Skipping AI models as requested)")
+            mapped_crops_df = pipeline.predict(
-        
+                unseen_pixel_df,
-        # We use dw_arr as the classification result
+                apply_spatial_smoothing=True,
-        cls_raster = dw_arr.copy()
+                coord_cols=['lat', 'lon']
+            )
             
             # ==========================================
-        # Stage 5: Apply Smoothing (Optional for DW)
+            # Stage 6: Export and Upload
             # ==========================================
-        if payload.get('smoothing_kernel'):
+            update_status(job_id, "running", "export_cog", 90, "Exporting results...")
-            kernel = payload['smoothing_kernel']
-            update_status(job_id, "running", "smooth", 75, f"Applying smoothing (k={kernel})...")
-            
-            from postprocess import majority_filter
-            
-            cls_raster = majority_filter(cls_raster, kernel=kernel, nodata=0)
-            print(f"[{job_id}] Smoothing applied")
-        
-        # ==========================================
-        # Stage 6: Export COGs
-        # ==========================================
-        update_status(job_id, "running", "export_cog", 80, "Exporting COGs...")
-        
-        from cog import write_cog
-        
             output_dir = Path(tempfile.mkdtemp())
+            output_path = output_dir / "refined.tif"
+            
+            pipeline.export_to_geotiff(mapped_crops_df, output_path=str(output_path))
+            
             output_urls = {}
-        missing_outputs = []
+            for filename in ["refined.tif", "refined_confidence.tif", "refined_cloud_mask.tif", "refined_legend.json"]:
+                local_f = output_dir / filename
+                if local_f.exists():
+                    result_key = f"results/{job_id}/{filename}"
+                    storage.upload_result(local_f, result_key)
+                    output_urls[filename.replace(".","_url")] = storage.presign_get("geocrop-results", result_key)
             
-        # Export refined raster
+        else:
-        if payload['outputs'].get('refined', True):
+            # Fallback to Legacy/DW-only logic (current implementation)
-            try:
+            print(f"[{job_id}] Using baseline logic (DW-only)...")
-                refined_path = output_dir / "refined.tif"
+            from dw_baseline import load_dw_baseline_window
-                dtype = "uint8" if cls_raster.max() <= 255 else "uint16"
+            # ... (keep existing Stage 3-6 logic for non-hybrid)
-                
-                write_cog(
-                    str(refined_path),
-                    cls_raster.astype(dtype),
-                    dw_profile,
-                    dtype=dtype,
-                    nodata=0,
-                )
-                
-                # Upload
-                result_key = f"results/{job_id}/refined.tif"
-                storage.upload_result(refined_path, result_key)
-                output_urls["refined_url"] = storage.presign_get("geocrop-results", result_key)
-                
-                print(f"[{job_id}] Exported refined.tif")
-                
-            except Exception as e:
-                missing_outputs.append(f"refined: {e}")
-        
-        # Export DW baseline if requested
-        if payload['outputs'].get('dw_baseline', False):
-            try:
-                dw_path = output_dir / "dw_baseline.tif"
-                write_cog(
-                    str(dw_path),
-                    dw_arr.astype("uint8"),
-                    dw_profile,
-                    dtype="uint8",
-                    nodata=0,
-                )
-                
-                result_key = f"results/{job_id}/dw_baseline.tif"
-                storage.upload_result(dw_path, result_key)
-                output_urls["dw_baseline_url"] = storage.presign_get("geocrop-results", result_key)
-                
-                print(f"[{job_id}] Exported dw_baseline.tif")
-                
-            except Exception as e:
-                missing_outputs.append(f"dw_baseline: {e}")
         
         # Note: indices and true_color not yet implemented
         if payload['outputs'].get('indices'):

k8s/base/kustomization.yaml

@@ -17,3 +17,4 @@ resources:
   - geocrop-web-ingress.yaml
   - geocrop-tiler-rewrite.yaml
   - 60-ingress-minio.yaml
+  - ntfy.yaml

k8s/base/ntfy.yaml

@@ -0,0 +1,99 @@
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: ntfy-data
+  namespace: monitoring
+spec:
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 5Gi
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: ntfy-config
+  namespace: monitoring
+data:
+  server.yml: |
+    base-url: "https://ntfy.techarvest.co.zw"
+    listen-http: ":80"
+    auth-file: "/var/lib/ntfy/user.db"
+    auth-default-access: "deny-all"
+    cache-file: "/var/lib/ntfy/cache.db"
+    attachment-cache-dir: "/var/lib/ntfy/attachments"
+---
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: ntfy
+  namespace: monitoring
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: ntfy
+  template:
+    metadata:
+      labels:
+        app: ntfy
+    spec:
+      containers:
+      - name: ntfy
+        image: binwiederhier/ntfy:latest
+        imagePullPolicy: Always
+        args: ["serve"]
+        ports:
+        - containerPort: 80
+        volumeMounts:
+        - name: ntfy-config
+          mountPath: /etc/ntfy/server.yml
+          subPath: server.yml
+          readOnly: true
+        - name: ntfy-data
+          mountPath: /var/lib/ntfy
+      volumes:
+      - name: ntfy-config
+        configMap:
+          name: ntfy-config
+      - name: ntfy-data
+        persistentVolumeClaim:
+          claimName: ntfy-data
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: ntfy
+  namespace: monitoring
+spec:
+  selector:
+    app: ntfy
+  ports:
+  - port: 80
+    targetPort: 80
+---
+apiVersion: networking.k8s.io/v1
+kind: Ingress
+metadata:
+  name: ntfy-ingress
+  namespace: monitoring
+  annotations:
+    cert-manager.io/cluster-issuer: "letsencrypt-prod"
+spec:
+  ingressClassName: nginx
+  tls:
+  - hosts:
+    - ntfy.techarvest.co.zw
+    secretName: ntfy-tls
+  rules:
+  - host: ntfy.techarvest.co.zw
+    http:
+      paths:
+      - path: /
+        pathType: Prefix
+        backend:
+          service:
+            name: ntfy
+            port:
+              number: 80