diff --git a/src/core/CL/CLHelpers.cpp b/src/core/CL/CLHelpers.cpp
index 1073b39..1c855e4 100644
--- a/src/core/CL/CLHelpers.cpp
+++ b/src/core/CL/CLHelpers.cpp
@@ -100,6 +100,32 @@
     }
 }
 
+std::string get_data_size_from_data_type(const DataType &dt)
+{
+    switch(dt)
+    {
+        case DataType::U8:
+        case DataType::QS8:
+        case DataType::S8:
+            return "8";
+        case DataType::U16:
+        case DataType::S16:
+        case DataType::QS16:
+        case DataType::F16:
+            return "16";
+        case DataType::U32:
+        case DataType::S32:
+        case DataType::F32:
+            return "32";
+        case DataType::U64:
+        case DataType::S64:
+            return "64";
+        default:
+            ARM_COMPUTE_ERROR("Unsupported input data type.");
+            return "0";
+    }
+}
+
 std::string get_underlying_cl_type_from_data_type(const DataType &dt)
 {
     switch(dt)
diff --git a/src/core/CL/CLKernelLibrary.cpp b/src/core/CL/CLKernelLibrary.cpp
index 9c8be36..dec2696 100644
--- a/src/core/CL/CLKernelLibrary.cpp
+++ b/src/core/CL/CLKernelLibrary.cpp
@@ -145,7 +145,8 @@
     { "copy_to_keypoint", "fast_corners.cl" },
     { "derivative", "derivative.cl" },
     { "dilate", "dilate.cl" },
-    { "direct_convolution3x3", "direct_convolution.cl" },
+    { "direct_convolution1x1", "direct_convolution1x1.cl" },
+    { "direct_convolution3x3", "direct_convolution3x3.cl" },
     { "erode", "erode.cl" },
     { "fast_corners", "fast_corners.cl" },
     { "fill_image_borders_constant", "fill_border.cl" },
@@ -350,8 +351,12 @@
 #include "./cl_kernels/dilate.clembed"
     },
     {
-        "direct_convolution.cl",
-#include "./cl_kernels/direct_convolution.clembed"
+        "direct_convolution1x1.cl",
+#include "./cl_kernels/direct_convolution1x1.clembed"
+    },
+    {
+        "direct_convolution3x3.cl",
+#include "./cl_kernels/direct_convolution3x3.clembed"
     },
     {
         "erode.cl",
diff --git a/src/core/CL/cl_kernels/direct_convolution1x1.cl b/src/core/CL/cl_kernels/direct_convolution1x1.cl
new file mode 100644
index 0000000..d161f80
--- /dev/null
+++ b/src/core/CL/cl_kernels/direct_convolution1x1.cl
@@ -0,0 +1,190 @@
+/*
+ * Copyright (c) 2016, 2017 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "helpers.h"
+
+#if STRIDE_X == 3
+#define INPUT_PIXEL_STR(data_size) extract_input_stride3_##data_size
+#define INPUT_PIXEL(data_size) INPUT_PIXEL_STR(data_size)
+#elif STRIDE_X == 2
+#define INPUT_PIXEL(data_size) extract_input_stride2
+#elif STRIDE_X == 1
+#define INPUT_PIXEL(data_size) extract_input_stride1
+#else /* STRIDE_X not equals 1, 2 or 3 */
+#error "Only support strides 1, 2 and 3"
+#endif /* STRIDE_X == 3 */
+
+/** Extracts a 1D horizontal vector from the input tensor with stride as 1.
+ *
+ * @param[in] input_pixel Pointer to the first pixel.
+ *
+ * @return extracted input pixels.
+ */
+inline VEC_DATA_TYPE(DATA_TYPE, 8) extract_input_stride1(__global const DATA_TYPE *input_pixel)
+{
+    return vload8(0, input_pixel);
+}
+
+/** Extracts a 1D horizontal vector from the input tensor with stride as 2.
+ *
+ * @param[in] input_pixel Pointer to the first pixel.
+ *
+ * @return extracted input pixels.
+ */
+inline VEC_DATA_TYPE(DATA_TYPE, 8) extract_input_stride2(__global const DATA_TYPE *input_pixel)
+{
+    VEC_DATA_TYPE(DATA_TYPE, 16)
+    temp = vload16(0, input_pixel);
+    return temp.s02468ace;
+}
+
+/** Extracts a 1D horizontal vector from the input tensor with stride as 3 and 32-bit data size.
+ *
+ * @param[in] input_pixel Pointer to the first pixel.
+ *
+ * @return extracted input pixels.
+ */
+inline VEC_DATA_TYPE(DATA_TYPE, 8) extract_input_stride3_32(__global const DATA_TYPE *input_pixel)
+{
+    VEC_DATA_TYPE(DATA_TYPE, 4)
+    temp1 = vload4(0, input_pixel);
+    VEC_DATA_TYPE(DATA_TYPE, 4)
+    temp2 = vload4(0, input_pixel + 6);
+    VEC_DATA_TYPE(DATA_TYPE, 4)
+    temp3 = vload4(0, input_pixel + 12);
+    VEC_DATA_TYPE(DATA_TYPE, 4)
+    temp4 = vload4(0, input_pixel + 18);
+    return (VEC_DATA_TYPE(DATA_TYPE, 8))(temp1.s03, temp2.s03, temp3.s03, temp4.s03);
+}
+
+/** Extracts a 1D horizontal vector from the input tensor with stride as 3 and 16-bit data size.
+ *
+ * @param[in] input_pixel Pointer to the first pixel.
+ *
+ * @return extracted input pixels.
+ */
+inline VEC_DATA_TYPE(DATA_TYPE, 8) extract_input_stride3_16(__global const DATA_TYPE *input_pixel)
+{
+    VEC_DATA_TYPE(DATA_TYPE, 8)
+    temp1 = vload8(0, input_pixel);
+    VEC_DATA_TYPE(DATA_TYPE, 8)
+    temp2 = vload8(0, input_pixel + 8);
+    VEC_DATA_TYPE(DATA_TYPE, 8)
+    temp3 = vload8(0, input_pixel + 16);
+    return (VEC_DATA_TYPE(DATA_TYPE, 8))(temp1.s036, temp2.s147, temp3.s25);
+}
+
+/** Extracts a 1D horizontal vector from the input tensor with stride as 3 and 8-bit data size.
+ *
+ * @param[in] input_pixel Pointer to the first pixel.
+ *
+ * @return extracted input pixels.
+ */
+inline VEC_DATA_TYPE(DATA_TYPE, 8) extract_input_stride3_8(__global const DATA_TYPE *input_pixel)
+{
+    VEC_DATA_TYPE(DATA_TYPE, 16)
+    temp1 = vload16(0, input_pixel);
+    VEC_DATA_TYPE(DATA_TYPE, 16)
+    temp2 = vload16(0, input_pixel + 12);
+    return (VEC_DATA_TYPE(DATA_TYPE, 8))(temp1.s0369, temp2.s0369);
+}
+
+/** This kernel performs a direct convolution to convolve the low three dimensions.
+ *
+ * @note The data type must be passed at compile time using -DDATA_TYPE: e.g. -DDATA_TYPE=float
+ * @note The data size must be passed at compile time using -DDATA_SIZE e.g. -DDATA_SIZE=32
+ * @note The convolution stride x and stride y must be passed at compile time using -DSTRIDE_X and -DSTRIDE_Y: e.g. -DSTRIDE_X=1, _DSTRIDE_Y=1
+ * @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
+ *
+ * @param[in]  src_ptr                               Pointer to the source tensor. Supported data types: QS8/F16/F32
+ * @param[in]  src_stride_x                          Stride of the source tensor in X dimension (in bytes)
+ * @param[in]  src_step_x                            src_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  src_stride_y                          Stride of the source tensor in Y dimension (in bytes)
+ * @param[in]  src_step_y                            src_stride_y * number of elements along Y processed per workitem(in bytes)
+ * @param[in]  src_stride_z                          Stride of the source tensor in Z dimension (in bytes)
+ * @param[in]  src_step_z                            src_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  src_offset_first_element_in_bytes     The offset of the first element in the source tensor
+ * @param[out] dst_ptr                               Pointer to the destination tensor. Supported data types: same as @p src_ptr
+ * @param[in]  dst_stride_x                          Stride of the destination tensor in X dimension (in bytes)
+ * @param[in]  dst_step_x                            dst_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  dst_stride_y                          Stride of the destination tensor in Y dimension (in bytes)
+ * @param[in]  dst_step_y                            dst_stride_y * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  dst_stride_z                          Stride of the destination tensor in Z dimension (in bytes)
+ * @param[in]  dst_step_z                            dst_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  dst_offset_first_element_in_bytes     The offset of the first element in the destination tensor
+ * @param[out] weights_ptr                           Pointer to the weights tensor. Supported data types: same as @p weights_ptr
+ * @param[in]  weights_stride_x                      Stride of the weights tensor in X dimension (in bytes)
+ * @param[in]  weights_step_x                        weights_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  weights_stride_y                      Stride of the weights tensor in Y dimension (in bytes)
+ * @param[in]  weights_step_y                        weights_stride_y * number of elements along y processed per workitem(in bytes)
+ * @param[in]  weights_stride_z                      Stride of the weights tensor in Z dimension (in bytes)
+ * @param[in]  weights_step_z                        weights_stride_z * number of elements along Z processed per workitem(in bytes)
+ * @param[in]  weights_offset_first_element_in_bytes The offset of the first element in the weights tensor
+ * @param[in]  biases_ptr                            Pointer to the biases tensor. Same as @p src_ptr
+ * @param[in]  biases_stride_x                       Stride of the biases tensor in X dimension (in bytes)
+ * @param[in]  biases_step_x                         biases_stride_x * number of elements along X processed per workitem(in bytes)
+ * @param[in]  biases_offset_first_element_in_bytes  The offset of the first element in the biases tensor
+ * @param[in]  weights_stride_w                      Stride of the weights tensor in W dimension
+ * @param[in]  filter_depth                          The depth size of the filter
+ */
+__kernel void direct_convolution1x1(
+    TENSOR3D_DECLARATION(src),
+    TENSOR3D_DECLARATION(dst),
+    TENSOR3D_DECLARATION(weights),
+#ifdef HAS_BIAS
+    VECTOR_DECLARATION(biases),
+#endif /* defined(HAS_BIAS) */
+    unsigned int weights_stride_w,
+    unsigned int filter_depth)
+{
+    Image    src     = CONVERT_TO_IMAGE_STRUCT(src);
+    Tensor3D weights = CONVERT_TO_TENSOR3D_STRUCT_NO_STEP(weights);
+    Tensor3D dst     = CONVERT_TO_TENSOR3D_STRUCT(dst);
+
+#ifdef HAS_BIAS
+    Vector biases = CONVERT_TO_VECTOR_STRUCT_NO_STEP(biases);
+#endif /* defined(HAS_BIAS) */
+
+    VEC_DATA_TYPE(DATA_TYPE, 8)
+    pixels = 0;
+
+    const uint z_index = get_global_id(2);
+
+    weights.ptr += z_index * weights_stride_w;
+
+    for(int d = 0; d < filter_depth; ++d)
+    {
+        DATA_TYPE weight = *(__global DATA_TYPE *)weights.ptr;
+        VEC_DATA_TYPE(DATA_TYPE, 8)
+        input_pixel = INPUT_PIXEL(DATA_SIZE)((__global DATA_TYPE *)src.ptr);
+        pixels += weight * input_pixel;
+        src.ptr += src_stride_z;
+        weights.ptr += weights_stride_z;
+    }
+
+#ifdef HAS_BIAS
+    pixels += (VEC_DATA_TYPE(DATA_TYPE, 8)) * ((__global DATA_TYPE *)(vector_offset(&biases, z_index)));
+#endif /* defined(HAS_BIAS) */
+
+    vstore8(pixels, 0, (__global DATA_TYPE *)dst.ptr);
+}
diff --git a/src/core/CL/cl_kernels/direct_convolution.cl b/src/core/CL/cl_kernels/direct_convolution3x3.cl
similarity index 100%
rename from src/core/CL/cl_kernels/direct_convolution.cl
rename to src/core/CL/cl_kernels/direct_convolution3x3.cl
diff --git a/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp b/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp
index 7f9e9d2..1f481de 100644
--- a/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp
+++ b/src/core/CL/kernels/CLDirectConvolutionLayerKernel.cpp
@@ -37,32 +37,33 @@
 
 using namespace arm_compute;
 
-template <unsigned int kernel_size>
-CLDirectConvolutionLayerKernel<kernel_size>::CLDirectConvolutionLayerKernel()
+CLDirectConvolutionLayerKernel::CLDirectConvolutionLayerKernel()
     : _input(nullptr), _biases(nullptr), _weights(nullptr), _output(nullptr), _border_size(0), _conv_pad_x(0), _conv_pad_y(0), _conv_stride_x(0), _conv_stride_y(0)
 {
 }
 
-template <unsigned int kernel_size>
-BorderSize             CLDirectConvolutionLayerKernel<kernel_size>::border_size() const
+BorderSize CLDirectConvolutionLayerKernel::border_size() const
 {
     return _border_size;
 }
 
-template <unsigned int kernel_size>
-void CLDirectConvolutionLayerKernel<kernel_size>::configure(const ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info)
+void CLDirectConvolutionLayerKernel::configure(const ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info)
 {
-    static_assert(kernel_size == 3, "Currently only 3x3 direct convolution is supported!");
-
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F32);
+    const unsigned int kernel_size = weights->info()->dimension(0);
+    ARM_COMPUTE_ERROR_ON_MSG(kernel_size != 1 && kernel_size != 3,
+                             "Kernel sizes other than 1x1 or 3x3 are not supported");
+    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16, DataType::F32);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, weights, output);
     ARM_COMPUTE_ERROR_ON(weights->info()->dimension(2) != input->info()->dimension(2));
     ARM_COMPUTE_ERROR_ON(weights->info()->dimension(0) != weights->info()->dimension(1));
     ARM_COMPUTE_ERROR_ON(weights->info()->num_dimensions() > 4);
+    ARM_COMPUTE_ERROR_ON_MSG(weights->info()->dimension(0) == 1 && (std::get<0>(conv_info.pad()) || std::get<1>(conv_info.pad())),
+                             "Pad > 0 not supported for 1x1 weights");
+    ARM_COMPUTE_ERROR_ON_MSG(weights->info()->dimension(0) == 3 && (std::get<0>(conv_info.pad()) > 1 || std::get<1>(conv_info.pad()) > 1),
+                             "Pad > 1 not supported for 3x3 weights");
+    ARM_COMPUTE_ERROR_ON_MSG(std::get<0>(conv_info.stride()) > 3, "Strides larger than 3 not supported.");
     ARM_COMPUTE_ERROR_ON_MSG((kernel_size == 3 && std::get<0>(conv_info.stride()) > 2), "Strides larger than 2 not supported in 3x3 direct convolution!");
 
-    ARM_COMPUTE_ERROR_ON(kernel_size != weights->info()->dimension(0));
-
     if(biases != nullptr)
     {
         ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(weights, biases);
@@ -86,6 +87,7 @@
     kernel_name << "direct_convolution" << kernel_size << "x" << kernel_size;
 
     options.insert("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type()));
+    options.insert("-DDATA_SIZE=" + get_data_size_from_data_type(input->info()->data_type()));
 
     options.emplace("-DSTRIDE_X=" + support::cpp11::to_string(_conv_stride_x));
 
@@ -130,8 +132,7 @@
     ICLKernel::configure(win);
 }
 
-template <unsigned int kernel_size>
-void CLDirectConvolutionLayerKernel<kernel_size>::run(const Window &window, cl::CommandQueue &queue)
+void CLDirectConvolutionLayerKernel::run(const Window &window, cl::CommandQueue &queue)
 {
     ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
     ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(IKernel::window(), window);
@@ -167,5 +168,3 @@
     }
     while(window.slide_window_slice_3D(slice) && win_in.slide_window_slice_3D(slice_in));
 }
-
-template class arm_compute::CLDirectConvolutionLayerKernel<3>;