Update num_classes parameter description in ObjectDetectionTask class (#2101)

* Update num_classes parameter description in ObjectDetectionTask class

* Update num_classes parameter description in SemanticSegmentationTask class

torchgeo/trainers/detection.py

@@ -82,7 +82,7 @@ class ObjectDetectionTask(BaseTask):
             weights: Initial model weights. True for ImageNet weights, False or None
                 for random weights.
             in_channels: Number of input channels to model.
-            num_classes: Number of prediction classes.
+            num_classes: Number of prediction classes (including the background).
             trainable_layers: Number of trainable layers.
             lr: Learning rate for optimizer.
             patience: Patience for learning rate scheduler.

torchgeo/trainers/segmentation.py

@@ -54,7 +54,7 @@ class SemanticSegmentationTask(BaseTask):
                 model does not support pretrained weights. Pretrained ViT weight enums
                 are not supported yet.
             in_channels: Number of input channels to model.
-            num_classes: Number of prediction classes.
+            num_classes: Number of prediction classes (including the background).
             num_filters: Number of filters. Only applicable when model='fcn'.
             loss: Name of the loss function, currently supports
                 'ce', 'jaccard' or 'focal' loss.