fix TSNE inference

utils/algo.py

@@ -600,12 +600,19 @@ class SKAlgorithm(IAMAPAlgorithm):
 
         ## some clustering algorithms need the entire dataset.
         do_fit_predict = False
+        do_fit_transform = False
 
         if not hasattr(model, "predict") and sk_module==cluster:
             do_fit_predict = True
 
+        if not hasattr(model, "transform") and sk_module==manifold:
+            do_fit_transform = True
+
         if do_fit_predict:
-            proj_img, model = self.fit_predict(model, feedback)
+            model = self.fit_predict(model, feedback)
+
+        if do_fit_transform:
+            model = self.fit_transform(model, feedback)
 
         else:
             iter = get_iter(model, fit_raster)
@@ -619,7 +626,7 @@ class SKAlgorithm(IAMAPAlgorithm):
             out_path = os.path.join(self.output_dir, save_file)
             joblib.dump(model, out_path)
 
-        if not do_fit_predict:
+        if not (do_fit_predict or do_fit_transform):
             feedback.pushInfo("Inference over raster\n")
             self.infer_model(model, feedback, scaler)
 
@@ -756,6 +763,52 @@ class SKAlgorithm(IAMAPAlgorithm):
         return model
 
 
+    def fit_transform(self, model, feedback):
+        with rasterio.open(self.rlayer_path) as ds:
+            transform = ds.transform
+            crs = ds.crs
+            win = windows.from_bounds(
+                self.extent.xMinimum(),
+                self.extent.yMinimum(),
+                self.extent.xMaximum(),
+                self.extent.yMaximum(),
+                transform=transform,
+            )
+            raster = ds.read(window=win)
+            transform = ds.window_transform(win)
+            raster = np.transpose(raster, (1, 2, 0))
+            raster = raster[:, :, self.input_bands]
+            fit_raster = raster.reshape(-1, raster.shape[-1])
+
+            # raster = (raster-np.mean(raster))/np.std(raster)
+            scaler = StandardScaler()
+
+            fit_raster = scaler.fit_transform(fit_raster)
+            np.nan_to_num(fit_raster)  # NaN to zero after normalisation
+
+            proj_img = model.fit_transform(fit_raster)
+
+            proj_img = proj_img.reshape((raster.shape[0], raster.shape[1], -1))
+            height, width, channels = proj_img.shape
+
+            feedback.pushInfo("Export to geotif\n")
+            with rasterio.open(
+                self.dst_path,
+                "w",
+                driver="GTiff",
+                height=height,
+                width=width,
+                count=channels,
+                dtype=self.out_dtype,
+                crs=crs,
+                transform=transform,
+            ) as dst_ds:
+                dst_ds.write(np.transpose(proj_img, (2, 0, 1)))
+            feedback.pushInfo("Export to geotif done\n")
+
+        return model
+
+
     def infer_model(self, model, feedback, scaler=None):
         with rasterio.open(self.rlayer_path) as ds:
             transform = ds.transform