add tf.pad() to pads a tensor.

Oceania2018 · Oceania2018 · commit 876275495e4e · 2019-08-30T20:26:27.000-05:00
diff --git a/src/TensorFlowNET.Core/APIs/tf.array.cs b/src/TensorFlowNET.Core/APIs/tf.array.cs
@@ -99,6 +99,18 @@ public Tensor one_hot(Tensor indices, int depth,
             int axis = -1,
             string name = null) => array_ops.one_hot(indices, depth, dtype: dtype, axis: axis, name: name);
 
+        /// <summary>
+        /// Pads a tensor
+        /// </summary>
+        /// <param name="tensor"></param>
+        /// <param name="paddings"></param>
+        /// <param name="mode"></param>
+        /// <param name="name"></param>
+        /// <param name="constant_values"></param>
+        /// <returns></returns>
+        public Tensor pad(Tensor tensor, Tensor paddings, string mode = "CONSTANT", string name = null, int constant_values = 0)
+            => array_ops.pad(tensor, paddings, mode: mode, name: name, constant_values: constant_values);
+
         /// <summary>
         /// A placeholder op that passes through `input` when its output is not fed.
         /// </summary>
diff --git a/src/TensorFlowNET.Core/Operations/array_ops.py.cs b/src/TensorFlowNET.Core/Operations/array_ops.py.cs
@@ -552,6 +552,40 @@ public static Tensor stack(object values, int axis = 0, string name = "stack")
             throw new NotImplementedException("array_ops.stack");
         }
 
+        public static Tensor pad(Tensor tensor, Tensor paddings, string mode = "CONSTANT", string name = null, int constant_values = 0)
+        {
+            Tensor result = null;
+            mode = mode.ToUpper();
+            if(mode == "CONSTANT")
+            {
+                if (constant_values != 0)
+                    throw new NotImplementedException("gen_array_ops.pad_v2");
+                else
+                    result = gen_array_ops.pad(tensor, paddings, name: name);
+            }
+
+            // Restore shape information where possible.
+            var paddings_constant = tensor_util.constant_value(
+                result.op.inputs[1], partial: true);
+            var input_shape = result.op.inputs[0].TensorShape;
+            if (input_shape.ndim > -1 &&
+                !result.TensorShape.is_fully_defined() &&
+                !(paddings_constant is null))
+            {
+                var new_shape = new List<int>();
+                foreach((NDArray padding, int dim) in zip(paddings_constant.GetNDArrays(), np.array(input_shape.dims).GetNDArrays()))
+                {
+                    if (padding is null || dim == -1 || padding.GetData<int>().Contains(-1))
+                        new_shape.Add(-1);
+                    else
+                        new_shape.Add(np.sum(padding) + dim);
+                }
+                result.set_shape(new_shape.ToArray());
+            }
+
+            return result;
+        }
+
         public static Tensor placeholder(TF_DataType dtype)
         {
             throw new NotImplementedException("array_ops.placeholder");
diff --git a/src/TensorFlowNET.Core/Operations/gen_array_ops.cs b/src/TensorFlowNET.Core/Operations/gen_array_ops.cs
@@ -99,6 +99,13 @@ public static Tensor gather_v2(Tensor @params, Tensor indices, int axis, string
             return _op.outputs[0];
         }
 
+        public static Tensor pad(Tensor input, Tensor paddings, string name = null)
+        {
+            var _op = _op_def_lib._apply_op_helper("Pad", name: name, args: new { input, paddings });
+
+            return _op.output;
+        }
+
         public static Tensor pack(Tensor[] values, int axis = 0, string name = null)
         {
             var _op = _op_def_lib._apply_op_helper("Pack", name: name, args: new { values, axis });
diff --git a/test/TensorFlowNET.Examples/ImageProcessing/YOLO/common.cs b/test/TensorFlowNET.Examples/ImageProcessing/YOLO/common.cs
@@ -19,6 +19,9 @@ public static Tensor convolutional(Tensor input_data, int[] filters_shape, Tenso
 
                 if (downsample)
                 {
+                    (int pad_h, int pad_w) = ((int)Math.Floor((filters_shape[0] - 2) / 2.0f) + 1, (int)Math.Floor((filters_shape[1] - 2) / 2.0f) + 1);
+                    var paddings = tf.constant(new int[,] { { 0, 0 }, { pad_h, pad_h }, { pad_w, pad_w }, { 0, 0 } });
+                    input_data = tf.pad(input_data, paddings, "CONSTANT");
                     throw new NotImplementedException("");
                 }
                 else

Original file line number	Diff line number	Diff line change
`@@ -99,6 +99,13 @@ public static Tensor gather_v2(Tensor @params, Tensor indices, int axis, string`
`99`	`99`	`return _op.outputs[0];`
`100`	`100`	`}`
`101`	`101`
	`102`	`+ public static Tensor pad(Tensor input, Tensor paddings, string name = null)`
	`103`	`+ {`
	`104`	`+ var _op = _op_def_lib._apply_op_helper("Pad", name: name, args: new { input, paddings });`
	`105`	`+`
	`106`	`+ return _op.output;`
	`107`	`+ }`
	`108`	`+`
`102`	`109`	`public static Tensor pack(Tensor[] values, int axis = 0, string name = null)`
`103`	`110`	`{`
`104`	`111`	`var _op = _op_def_lib._apply_op_helper("Pack", name: name, args: new { values, axis });`
Original file line number	Diff line number	Diff line change
`@@ -19,6 +19,9 @@ public static Tensor convolutional(Tensor input_data, int[] filters_shape, Tenso`
`19`	`19`
`20`	`20`	`if (downsample)`
`21`	`21`	`{`
	`22`	`+ (int pad_h, int pad_w) = ((int)Math.Floor((filters_shape[0] - 2) / 2.0f) + 1, (int)Math.Floor((filters_shape[1] - 2) / 2.0f) + 1);`
	`23`	`+ var paddings = tf.constant(new int[,] { { 0, 0 }, { pad_h, pad_h }, { pad_w, pad_w }, { 0, 0 } });`
	`24`	`+ input_data = tf.pad(input_data, paddings, "CONSTANT");`
`22`	`25`	`throw new NotImplementedException("");`
`23`	`26`	`}`
`24`	`27`	`else`