diff --git a/python/GafferMLUI/ImageToTensorUI.py b/python/GafferMLUI/ImageToTensorUI.py
index f9fc11ce50d..b932ac870d9 100644
--- a/python/GafferMLUI/ImageToTensorUI.py
+++ b/python/GafferMLUI/ImageToTensorUI.py
@@ -41,19 +41,76 @@
 
 	GafferML.ImageToTensor,
 
-	# "description",
-	# """
-	# Converts Gaffer data to tensors for use with the Inference node.
-	# Potential data sources include PrimitiveVariableQuery nodes to fetch data
-	# from 3D scenes, or expressions to generate arbitrary input data.
-	# """,
+	"description",
+	"""
+	Converts images to tensors for use with the Inference node.
+
+	> Note : Only the data window is converted, as it would typically be
+	> wasteful to convert and process the empty pixels outside the data window.
+	> If this is necessary, merge the image over a Constant image before
+	> conversion.
+	""",
 
 	plugs = {
 
+		"image" : [
+
+			"description",
+			"""
+			The image to be converted.
+			""",
+
+		],
+
 		"view" : [
 
+			"description",
+			"""
+			The image view to take the tensor data from.
+			""",
+
 			"plugValueWidget:type", "GafferImageUI.ViewPlugValueWidget",
 
+			"noduleLayout:visible", False,
+
+		],
+
+		"channels" : [
+
+			"description",
+			"""
+			The list of channels to convert. Channels are added to the
+			tensor in the order specified, so can be shuffled by changing
+			the order. For example, an order of `[ "B", "G", "R" ]` might
+			be needed for use with model trained on images using OpenCV
+			conventions.
+			""",
+
+			"noduleLayout:visible", False,
+
+		],
+
+		"interleaveChannels" : [
+
+			"description",
+			"""
+			Interleaves the channel data, so that all channels for a single
+			pixel are adjacent in memory.
+			""",
+
+			"noduleLayout:visible", False,
+
+		],
+
+		"tensor" : [
+
+			"description",
+			"""
+			The output tensor.
+			""",
+
+			"layout:visibilityActivator", lambda plug : False,
+
 		],
 
 	}