Grouping layer added

dlib/dnn/core.h

@@ -3159,6 +3159,499 @@ namespace dlib
 
 // ----------------------------------------------------------------------------------------
 
+    namespace impl
+    {
+        template <typename T>
+        struct group_helper;
+        template<typename... R>
+        struct group_count_helper;
+    }
+
+
+    // --------------------------------------------------------------------------------------
+    // this class is used to reference group layer input
+    class group_input
+    {
+    public:
+        typedef tensor input_type;
+        const static unsigned int sample_expansion_factor = 1;
+        friend void serialize(const group_input& item, std::ostream& out)
+        {
+            serialize("group_input", out);
+        }
+
+        friend void deserialize(group_input& item, std::istream& in)
+        {
+            std::string version;
+            deserialize(version, in);
+            if (version != "group_input")
+                throw serialization_error("Unexpected version found while deserializing dlib::group_input.");
+        }
+
+        friend std::ostream& operator<<(std::ostream& out, const group_input& item)
+        {
+            out << "group_input";
+            return out;
+        }
+    };
+    // --------------------------------------------------------------------------------------
+
+    template <typename GRP, typename SUBNET>
+    class depth_group;
+
+
+    template <typename T, typename U>
+    struct is_nonloss_layer_type<depth_group<T,U>> : std::true_type {};
+
+    template <typename GRP, typename SUBNET>
+    class depth_group
+    {
+    public:
+        typedef GRP grp_type;
+        typedef SUBNET subnet_type;
+        typedef typename subnet_type::input_type input_type;
+        const static size_t group_size = std::tuple_size<grp_type>::value;
+        const static size_t num_layers_in_group = impl::group_count_helper<GRP>::num_layers;
+        const static size_t num_layers = subnet_type::num_layers + num_layers_in_group;
+        const static size_t num_computational_layers_in_group = impl::group_count_helper<GRP>::num_computational_layers;
+        const static size_t num_computational_layers = subnet_type::num_computational_layers + num_computational_layers_in_group;
+        const static unsigned int sample_expansion_factor = subnet_type::sample_expansion_factor;
+
+        using group_helper = impl::group_helper<grp_type>;
+
+        depth_group(
+        ):
+                subnetwork(new subnet_type()),
+                grp(new grp_type()),
+                gradient_input_is_stale(true),
+                get_output_and_gradient_input_disabled(false)
+        {
+        }
+
+        depth_group(const depth_group& item)
+        {
+            grp.reset(new grp_type(*item.grp));
+            subnetwork.reset(new subnet_type(*item.subnetwork));
+            gradient_input_is_stale = item.gradient_input_is_stale;
+            get_output_and_gradient_input_disabled = item.get_output_and_gradient_input_disabled;
+            x_grad = item.x_grad;
+            cached_output = item.cached_output;
+            temp_tensor = item.temp_tensor;
+        }
+        depth_group& operator=(const depth_group& item) { depth_group(item).swap(*this); return *this;}
+        depth_group(depth_group&& item) : depth_group() { swap(item); }
+        depth_group& operator=(depth_group&& item) { swap(item); return *this; }
+
+        template <typename T, typename U, typename E>
+        friend class add_layer;
+        template <typename T, bool is_first, typename E>
+        friend class dimpl::subnet_wrapper;
+        template <unsigned long T, typename U, typename E>
+        friend class add_tag_layer;
+        template <template<typename> class T, typename U>
+        friend class add_skip_layer;
+        template <size_t N, template<typename> class L, typename S>
+        friend class repeat;
+
+        // Allow copying networks from one to another as long as their corresponding
+        // layers can be constructed from each other.
+        template <typename T, typename U>
+        depth_group(
+                const depth_group<T,U>& item
+        ) :
+                grp(new grp_type(item.detail())),
+                subnetwork(new subnet_type(item.subnet())),
+                gradient_input_is_stale(item.gradient_input_is_stale),
+                get_output_and_gradient_input_disabled(item.get_output_and_gradient_input_disabled),
+                x_grad(item.x_grad),
+                cached_output(item.cached_output)
+        {
+        }
+
+        template <typename input_iterator>
+        void to_tensor (
+                input_iterator ibegin,
+                input_iterator iend,
+                resizable_tensor& data
+        ) const
+        {
+            subnetwork->to_tensor(ibegin,iend,data);
+        }
+
+        template <typename input_iterator>
+        const tensor& operator() (
+                input_iterator ibegin,
+                input_iterator iend
+        )
+        {
+            to_tensor(ibegin,iend,temp_tensor);
+            return forward(temp_tensor);
+        }
+
+
+        const tensor& operator() (const input_type& x)
+        {
+            return (*this)(&x, &x+1);
+        }
+
+
+        // forward for group: subnet->for_each_in_group->concat->cached_output
+        const tensor& forward(const tensor& x)
+        {
+
+            subnetwork->forward(x);
+            long group_depth = 0;
+
+            group_helper::forward(subnetwork->get_output(), detail(), group_depth);
+
+            auto& out_0 = std::get<0>(detail()).get_output();
+            cached_output.set_size(out_0.num_samples(), group_depth, out_0.nr(), out_0.nc());
+
+            group_helper::concat(cached_output, detail());
+
+
+            gradient_input_is_stale = true;
+            return private_get_output();
+        }
+
+    private:
+        bool this_layer_requires_forward_output(
+        )
+        {
+            return true;
+        }
+
+        tensor& private_get_output() const
+        {
+            return const_cast<resizable_tensor&>(cached_output);
+        }
+        tensor& private_get_gradient_input()
+        {
+            if (gradient_input_is_stale)
+            {
+                gradient_input_is_stale = false;
+                x_grad.copy_size(private_get_output());
+                x_grad = 0;
+            }
+            return x_grad;
+        }
+        void disable_output_and_gradient_getters (
+        ) { get_output_and_gradient_input_disabled = true; }
+    public:
+        const tensor& get_output() const
+        {
+            if (get_output_and_gradient_input_disabled)
+                throw dlib::error("Accessing this layer's get_output() is disabled because an in-place layer has been stacked on top of it.");
+            return private_get_output();
+        }
+        tensor& get_gradient_input()
+        {
+            if (get_output_and_gradient_input_disabled)
+                throw dlib::error("Accessing this layer's get_gradient_input() is disabled because an in-place layer has been stacked on top of it.");
+            return private_get_gradient_input();
+        }
+
+        const tensor& get_final_data_gradient(
+        ) const { return subnetwork->get_final_data_gradient(); }
+
+        void back_propagate_error(const tensor& x)
+        {
+            back_propagate_error(x, private_get_gradient_input());
+        }
+        void back_propagate_error(const tensor& x, const tensor& gradient_input)
+        {
+            group_helper::backward(detail(), get_gradient_input(), subnetwork->get_output(), subnetwork->get_gradient_input());
+
+            subnetwork->back_propagate_error(x);
+
+            // zero out get_gradient_input()
+            gradient_input_is_stale = true;
+        }
+
+        template <typename solver_type>
+        void update_parameters(sstack<solver_type> solvers, double step_size)
+        {
+            DLIB_CASSERT(solvers.size()>=num_computational_layers,"");
+            group_helper::update_parameters(solvers, step_size, detail());
+            solvers = solvers.pop(num_computational_layers_in_group);
+            subnetwork->update_parameters(solvers, step_size);
+        }
+
+        const subnet_type& subnet() const { return *subnetwork; }
+        subnet_type& subnet() { return *subnetwork; }
+
+        const grp_type& detail() const { return *grp; }
+        grp_type& detail() { return *grp; }
+
+        void clean()
+        {
+            x_grad.clear();
+            cached_output.clear();
+            temp_tensor.clear();
+            gradient_input_is_stale = true;
+            subnetwork->clean();
+        }
+
+        friend void serialize(const depth_group& item, std::ostream& out)
+        {
+            int version = 2;
+            serialize(version, out);
+            serialize(*item.subnetwork, out);
+            group_helper::serialize(*item.grp, out);
+            serialize(item.gradient_input_is_stale, out);
+            serialize(item.get_output_and_gradient_input_disabled, out);
+            serialize(item.x_grad, out);
+            serialize(item.cached_output, out);
+        }
+
+        friend void deserialize(depth_group& item, std::istream& in)
+        {
+            int version = 0;
+            deserialize(version, in);
+            if (!(1 <= version && version <= 2))
+                throw serialization_error("Unexpected version found while deserializing dlib::depth_group.");
+            deserialize(*item.subnetwork, in);
+            group_helper::deserialize(*item.grp, in);
+            deserialize(item.gradient_input_is_stale, in);
+            deserialize(item.get_output_and_gradient_input_disabled, in);
+            deserialize(item.x_grad, in);
+            deserialize(item.cached_output, in);
+        }
+
+        friend std::ostream& operator<< (std::ostream& out, const depth_group& item)
+        {
+            item.print(out, 0);
+            return out;
+        }
+
+        void print (std::ostream& out, unsigned long idx=0) const
+        {
+            out << "layer<" << idx << ">\t";
+            detail().print(out, idx);
+            subnet().print(out, idx+1);
+        }
+
+    private:
+
+
+        void swap(depth_group& item)
+        {
+            std::swap(subnetwork,item.subnetwork);
+            std::swap(grp, item.grp);
+            std::swap(gradient_input_is_stale, item.gradient_input_is_stale);
+            std::swap(get_output_and_gradient_input_disabled, item.get_output_and_gradient_input_disabled);
+            std::swap(x_grad, item.x_grad);
+            std::swap(cached_output, item.cached_output);
+        }
+
+
+        std::unique_ptr<subnet_type> subnetwork;
+        std::unique_ptr<grp_type> grp;
+
+        bool gradient_input_is_stale;
+        bool get_output_and_gradient_input_disabled;
+
+        resizable_tensor x_grad;
+        resizable_tensor cached_output;
+
+        // temp_tensor doesn't logically contribute to the state of this object.
+        // It is here only to prevent it from being reallocated over and over.
+        resizable_tensor temp_tensor;
+    };
+
+    // define "grp" layer shorter name for usage when creating networks
+    template <typename GRP, typename SUBNET>
+    using grp = depth_group<GRP, SUBNET>;
+
+    namespace impl {
+        template<
+                unsigned int i,
+                typename T, typename U
+        >
+        struct layer_helper<i, depth_group<T, U>,
+                typename std::enable_if<(i != 0 && i >= depth_group<T, U>::num_layers_in_group)>::type> {
+            const static size_t num_layers_in_group = depth_group<T, U>::num_layers_in_group;
+
+            using next_type = typename depth_group<T, U>::subnet_type;
+            using type = typename layer_helper<i - num_layers_in_group, next_type>::type;
+
+            static type &layer(depth_group<T, U> &n) {
+                return layer_helper<i - num_layers_in_group, next_type>::layer(n.subnet());
+            }
+        };
+
+        template<
+                unsigned int i,
+                typename T, typename U
+        >
+        struct layer_helper<i, depth_group<T, U>,
+                typename std::enable_if<(i != 0 && i < depth_group<T, U>::num_layers_in_group)>::type> {
+            const static size_t num_layers_in_group = depth_group<T, U>::num_layers_in_group;
+            typedef typename depth_group<T, U>::grp_type grp_type;
+            using type = typename layer_helper<i, grp_type>::type;
+
+            static type &layer(depth_group<T, U> &n) {
+                return layer_helper<i, grp_type>::layer(n.detail());
+            }
+        };
+
+        template <unsigned int pos, unsigned int i, typename... T>
+        struct group_pos_search{
+            const static unsigned int count = sizeof...(T);
+            const static unsigned int pos_from_begin = count - pos - 1;
+            using tuple_elem_type = typename std::tuple_element<pos_from_begin, std::tuple<T...>>::type;
+            static const unsigned int num_layers = tuple_elem_type::num_layers;
+
+            static const unsigned int layer_index = i >= num_layers ? group_pos_search<pos - 1, i - num_layers, T...>::layer_index : i;
+            static const unsigned int tuple_index = i >= num_layers ? group_pos_search<pos - 1, i - num_layers, T...>::tuple_index + 1 : pos;
+        };
+        template <unsigned int i, typename... T>
+        struct group_pos_search<0, i, T...>{
+            static const unsigned int layer_index = i;
+            static const unsigned int tuple_index = 0;
+        };
+
+
+        template<
+                unsigned int i,
+                typename... R
+        >
+        struct layer_helper<i, std::tuple<R...>, typename std::enable_if<true>::type>{
+            const static unsigned tuple_size = sizeof...(R);
+
+            static const unsigned int layer_index = group_pos_search<tuple_size - 1, i, R...>::layer_index;
+            static const unsigned int tuple_index = group_pos_search<tuple_size - 1, i, R...>::tuple_index;
+
+            using next_type = typename std::tuple_element<tuple_index, std::tuple<R...>>::type;//typename std::remove_reference<decltype(makeT().subnet())>::type;
+            using type = typename layer_helper<layer_index,next_type>::type;
+
+            static type &layer(std::tuple<R...> &n) {
+                return layer_helper<layer_index, next_type>::layer(std::get<tuple_index>(n));
+            }
+        };
+
+        // helper classes for layer group processing
+        template <size_t idx, typename... T>
+        struct group_helper_impl{
+            static void serialize_impl(const std::tuple<T...>& data, std::ostream& out){
+                group_helper_impl<idx - 1, T...>::serialize_impl(data, out);
+                serialize(std::get<idx>(data), out);
+            }
+            static void deserialize_impl(std::tuple<T...>& data, std::istream& in){
+                group_helper_impl<idx - 1, T...>::deserialize_impl(data, in);
+                deserialize(std::get<idx>(data), in);
+            }
+            static void forward(const tensor& x, std::tuple<T...>& grp, long& group_depth){
+                group_helper_impl<idx - 1, T...>::forward(x, grp, group_depth);
+                auto& r = std::get<idx>(grp).forward(x);
+                group_depth += r.k();
+            }
+            static size_t concat(resizable_tensor& cached_output, std::tuple<T...>& grp, size_t offset){
+                offset += group_helper_impl<idx - 1, T...>::concat(cached_output, grp, offset);
+                auto& output = std::get<idx>(grp).get_output();
+                tt::concat_depth(cached_output, offset, output);
+                return offset + output.nc() * output.nr() * output.k();
+            }
+            template<typename solver_type>
+            static sstack<solver_type> update_parameters(sstack<solver_type> solvers, double step_size, std::tuple<T...>& grp){
+                sstack<solver_type> sub_solvers = group_helper_impl<idx - 1, T...>::update_parameters(solvers, step_size, grp);
+                std::get<idx>(grp).update_parameters(sub_solvers, step_size);
+                using tuple_elem_type = typename std::tuple_element<idx, std::tuple<T...>>::type;
+                return sub_solvers.pop(tuple_elem_type::num_computational_layers);
+            }
+            static size_t backward(std::tuple<T...>& grp, const tensor& group_gradient_in,
+                                         const tensor& subnet_out, tensor& group_gradient_out, size_t offset)
+            {
+                offset += group_helper_impl<idx - 1, T...>::backward(grp, group_gradient_in, subnet_out, group_gradient_out, offset);
+
+                auto& subnet = std::get<idx>(grp);
+                auto& gr_input = subnet.get_gradient_input();
+                tt::split_depth(gr_input, offset, group_gradient_in);
+
+                subnet.back_propagate_error(subnet_out);
+
+                tt::add(group_gradient_out, group_gradient_out, subnet.get_final_data_gradient());
+                return offset + gr_input.nc() * gr_input.nr() * gr_input.k();
+            }
+        };
+        template <typename... T>
+        struct group_helper_impl<0, T...>{
+            static void serialize_impl(const std::tuple<T...>& data, std::ostream& out){
+                serialize(std::get<0>(data), out);
+            }
+            static void deserialize_impl(std::tuple<T...>& data, std::istream& in){
+                deserialize(std::get<0>(data), in);
+            }
+            static void forward(const tensor& x, std::tuple<T...>& grp, long& group_depth){
+                auto& r = std::get<0>(grp).forward(x);
+                group_depth += r.k();
+            }
+            static size_t concat(resizable_tensor& cached_output, std::tuple<T...>& grp, size_t offset){
+                auto& output = std::get<0>(grp).get_output();
+                tt::concat_depth(cached_output, offset, output);
+                return offset + output.nc() * output.nr() * output.k();
+            }
+            template<typename solver_type>
+            static sstack<solver_type> update_parameters(sstack<solver_type> solvers, double step_size, std::tuple<T...>& grp){
+                std::get<0>(grp).update_parameters(solvers, step_size);
+                using tuple_elem_type = typename std::tuple_element<0, std::tuple<T...>>::type;
+                return solvers.pop(tuple_elem_type::num_computational_layers);
+            }
+            static size_t backward(std::tuple<T...>& grp, const tensor& group_gradient_in,
+                                         const tensor& subnet_out, tensor& group_gradient_out, size_t offset)
+            {
+                auto& item = std::get<0>(grp);
+                auto& gr_input = item.get_gradient_input();
+                tt::split_depth(gr_input, offset, group_gradient_in);
+                item.back_propagate_error(subnet_out);
+
+                tt::add(group_gradient_out, group_gradient_out, item.get_final_data_gradient());
+                return offset + gr_input.nc() * gr_input.nr() * gr_input.k();
+            }
+        };
+        template <typename... T>
+        struct group_helper<std::tuple<T...>>{
+            static void serialize(const std::tuple<T...> & data, std::ostream& out){
+                group_helper_impl<std::tuple_size<std::tuple<T...>>::value - 1, T...>::serialize_impl(data, out);
+            }
+            static void deserialize(std::tuple<T...>& data, std::istream& in){
+                group_helper_impl<std::tuple_size<std::tuple<T...>>::value - 1, T...>::deserialize_impl(data, in);
+            }
+            static void forward(const tensor& x, std::tuple<T...>& grp, long& group_depth){
+                group_helper_impl<std::tuple_size<std::tuple<T...>>::value - 1, T...>::forward(x, grp, group_depth);
+            }
+            static void concat(resizable_tensor& out, std::tuple<T...>& grp){
+                group_helper_impl<std::tuple_size<std::tuple<T...>>::value - 1, T...>::concat(out, grp, 0);
+            }
+            template<typename solver_type>
+            static void update_parameters(sstack<solver_type> solvers, double step_size, std::tuple<T...>& grp){
+                group_helper_impl<std::tuple_size<std::tuple<T...>>::value - 1, T...>::update_parameters(solvers, step_size, grp);
+            }
+            static void backward(std::tuple<T...>& grp, const tensor& group_gradient_in, const tensor& subnet_out, tensor& group_gradient_out)
+            {
+                group_helper_impl<std::tuple_size<std::tuple<T...>>::value - 1, T...>::backward(grp, group_gradient_in, subnet_out, group_gradient_out, 0);
+            }
+        };
+
+        // helper classes to understand the count of group items layers
+        template<typename T>
+        struct group_count_helper<T>{
+            const static size_t num_layers = T::num_layers;
+            const static size_t num_computational_layers = T::num_computational_layers;
+        };
+
+        template<typename T, typename... R>
+        struct group_count_helper<T, R...>{
+            const static size_t num_layers = group_count_helper<T>::num_layers + group_count_helper<R...>::num_layers;
+            const static size_t num_computational_layers = group_count_helper<T>::num_computational_layers + group_count_helper<R...>::num_computational_layers;
+        };
+        template<typename... R>
+        struct group_count_helper<std::tuple<R...>>{
+            const static size_t num_layers = group_count_helper<R...>::num_layers;
+            const static size_t num_computational_layers = group_count_helper<R...>::num_computational_layers;
+        };
+
+    }
 }
 
 #endif // DLIB_DNn_CORE_H_

dlib/dnn/tensor_tools.cpp

@@ -634,6 +634,66 @@ namespace dlib { namespace tt
 #endif
     }
 
+    // ----------------------------------------------------------------------------------------
+    // ------------------------------------------------------------------------------------
+
+    void concat_depth(tensor& dest, size_t sample_offset, const tensor& src)
+    {
+        const size_t dest_sample_size = static_cast<size_t>(dest.nc() * dest.nr() * dest.k());
+        const size_t src_sample_size = static_cast<size_t>(src.nc() * src.nr() * src.k());
+
+        DLIB_CASSERT(dest.num_samples() == src.num_samples() &&
+                     dest.nc() == src.nc() && dest.nr() == src.nr(), "All sources should fit into dest tensor size");
+        DLIB_CASSERT(dest_sample_size >= src_sample_size + sample_offset, "Not enough space in dest tensor");
+
+#ifdef DLIB_USE_CUDA
+        float* dest_p = dest.device_write_only() + sample_offset;
+        const float* src_p = src.device();
+#else
+        float* dest_p = dest.host_write_only() + sample_offset;
+        const float* src_p = src.host();
+#endif
+
+        for (unsigned long i = 0; i < src.num_samples(); ++i)
+        {
+#ifdef DLIB_USE_CUDA
+            CHECK_CUDA(cudaMemcpy(dest_p, src_p, src_sample_size * sizeof(float), cudaMemcpyDeviceToDevice));
+#else
+            ::memcpy(dest_p, src_p, src_sample_size * sizeof(float));
+#endif
+            dest_p += dest_sample_size;
+            src_p  += src_sample_size;
+        }
+    }
+
+    void split_depth(tensor& dest, size_t sample_offset, const tensor& src)
+    {
+        const size_t dest_sample_size = static_cast<size_t>(dest.nc() * dest.nr() * dest.k());
+        const size_t src_sample_size = static_cast<size_t>(src.nc() * src.nr() * src.k());
+
+        DLIB_CASSERT(dest.num_samples() == src.num_samples() &&
+                     dest.nc() == src.nc() && dest.nr() == src.nr(), "All sources should fit into dest tensor size");
+        DLIB_CASSERT(dest_sample_size <= src_sample_size - sample_offset, "Not enough space in dest tensor");
+
+#ifdef DLIB_USE_CUDA
+        float* dest_p = dest.device_write_only();
+        const float* src_p = src.device() + sample_offset;
+#else
+        float* dest_p = dest.host_write_only();
+        const float* src_p = src.host() + sample_offset;
+#endif
+
+        for (unsigned long i = 0; i < src.num_samples(); ++i)
+        {
+#ifdef DLIB_USE_CUDA
+            CHECK_CUDA(cudaMemcpy(dest_p, src_p, dest_sample_size * sizeof(float), cudaMemcpyDeviceToDevice));
+#else
+            ::memcpy(dest_p, src_p, dest_sample_size * sizeof(float));
+#endif
+            dest_p += dest_sample_size;
+            src_p  += src_sample_size;
+        }
+    }
 // ----------------------------------------------------------------------------------------
 
 }}

dlib/dnn/tensor_tools.h

@@ -1171,6 +1171,43 @@ namespace dlib { namespace tt
 
         resizable_tensor accum_buffer;
     };
+    // ----------------------------------------------------------------------------------------
+
+    void concat_depth(
+            tensor& dest,
+            size_t sample_offset,
+            const tensor& src
+    );
+    /*!
+        requires
+            - dest.nc() == src.nc()
+            - dest.nr() == src.nr()
+            - dest.num_samples() == src.num_samples()
+            - dest.k() >= src.k() + sample_offset
+            - is_same_object(dest,src) == false
+            - sample_offset a count of elements, not bytes
+        ensures
+            - performs: dest[i, k + sample_offset, r, c] = src[i, k, r, c], where k in [0..src.k()]
+              Copies content of each sample from src in to corresponding place of sample at dst
+    !*/
+
+    void split_depth(
+            tensor& dest,
+            size_t sample_offset,
+            const tensor& src
+    );
+    /*!
+        requires
+            - dest.nc() == src.nc()
+            - dest.nr() == src.nr()
+            - dest.num_samples() == src.num_samples()
+            - dest.k() <= src.k() - sample_offset
+            - is_same_object(dest,src) == false
+            - sample_offset a count of elements, not bytes
+        ensures
+            - performs: dest[i, k, r, c] = src[i, k  + sample_offset, r, c], where k in [0..dest.k()]
+              Fills each sample of dst from the corresponding part of each sample at src
+    !*/
 
 // ----------------------------------------------------------------------------------------
 

examples/CMakeLists.txt

@@ -33,6 +33,7 @@ ENDMACRO()
 if (COMPILER_CAN_DO_CPP_11)
    add_example(dnn_mnist_ex)
    add_example(dnn_mnist_advanced_ex)
+   add_example(dnn_inception_ex)
 endif()
 
 #here we apply our macros 

examples/dnn_inception_ex.cpp

+// The contents of this file are in the public domain. See LICENSE_FOR_EXAMPLE_PROGRAMS.txt
+/*
+    This is an example illustrating the use of the deep learning tools from the
+    dlib C++ Library.  I'm assuming you have already read the dnn_mnist_ex.cpp
+    example.  So in this example program I'm going to go over a number of more
+    advanced parts of the API, including:
+        - Using grp layer for constructing inception layer
+
+    Inception layer is a kind of NN architecture for running sevelar convolution types
+    on the same input area and joining all convolution results into one output.
+    For further reading refer http://www.cs.unc.edu/~wliu/papers/GoogLeNet.pdf
+*/
+
+
+#include <dlib/dnn.h>
+#include <iostream>
+#include <dlib/data_io.h>
+#include <tuple>
+
+using namespace std;
+using namespace dlib;
+
+// Here we define inception module as described in GoogLeNet specification. The depth of each sublayer can be changed
+template<typename SUBNET>
+using inception = grp<std::tuple<con<8,1,1,1,1, group_input>,
+                                 con<8,3,3,1,1, con<8,1,1,1,1, group_input>>,
+                                 con<8,5,5,1,1, con<8,1,1,1,1, group_input>>,
+                                 con<8,1,1,1,1, max_pool<3,3,1,1, group_input>>>,
+                      SUBNET>;
+
+int main(int argc, char** argv) try
+{
+    // This example is going to run on the MNIST dataset.  
+    if (argc != 2)
+    {
+        cout << "This example needs the MNIST dataset to run!" << endl;
+        cout << "You can get MNIST from http://yann.lecun.com/exdb/mnist/" << endl;
+        cout << "Download the 4 files that comprise the dataset, decompress them, and" << endl;
+        cout << "put them in a folder.  Then give that folder as input to this program." << endl;
+        return 1;
+    }
+
+
+    std::vector<matrix<unsigned char>> training_images;
+    std::vector<unsigned long>         training_labels;
+    std::vector<matrix<unsigned char>> testing_images;
+    std::vector<unsigned long>         testing_labels;
+    load_mnist_dataset(argv[1], training_images, training_labels, testing_images, testing_labels);
+
+
+    // Create a the same network as in dnn_mnist_ex, but use inception layer insteam of convolution
+    // in the middle
+    using net_type = loss_multiclass_log<
+            fc<10,
+                    relu<fc<84,
+                            relu<fc<120,
+                                    max_pool<2,2,2,2,relu<inception<
+                                            max_pool<2,2,2,2,relu<con<6,5,5,1,1,
+                                                    input<matrix<unsigned char>>
+                                            >>>>>>>>>>>>;
+
+
+    // Create a network as defined above.  This network will produce 10 outputs
+    // because that's how we defined net_type.  However, fc layers can have the
+    // number of outputs they produce changed at runtime.
+    net_type net;
+
+    // the following training process is the same as in dnn_mnist_ex sample
+
+    // And then train it using the MNIST data.  The code below uses mini-batch stochastic
+    // gradient descent with an initial learning rate of 0.01 to accomplish this.
+    dnn_trainer<net_type> trainer(net);
+    trainer.set_learning_rate(0.01);
+    trainer.set_min_learning_rate(0.00001);
+    trainer.set_mini_batch_size(128);
+    trainer.be_verbose();
+    // Since DNN training can take a long time, we can ask the trainer to save its state to
+    // a file named "mnist_sync" every 20 seconds.  This way, if we kill this program and
+    // start it again it will begin where it left off rather than restarting the training
+    // from scratch.  This is because, when the program restarts, this call to
+    // set_synchronization_file() will automatically reload the settings from mnist_sync if
+    // the file exists.
+    trainer.set_synchronization_file("mnist_sync", std::chrono::seconds(20));
+    // Finally, this line begins training.  By default, it runs SGD with our specified
+    // learning rate until the loss stops decreasing.  Then it reduces the learning rate by
+    // a factor of 10 and continues running until the loss stops decreasing again.  It will
+    // keep doing this until the learning rate has dropped below the min learning rate
+    // defined above or the maximum number of epochs as been executed (defaulted to 10000). 
+    trainer.train(training_images, training_labels);
+
+    // At this point our net object should have learned how to classify MNIST images.  But
+    // before we try it out let's save it to disk.  Note that, since the trainer has been
+    // running images through the network, net will have a bunch of state in it related to
+    // the last batch of images it processed (e.g. outputs from each layer).  Since we
+    // don't care about saving that kind of stuff to disk we can tell the network to forget
+    // about that kind of transient data so that our file will be smaller.  We do this by
+    // "cleaning" the network before saving it.
+    net.clean();
+    serialize("mnist_network.dat") << net;
+    // Now if we later wanted to recall the network from disk we can simply say:
+    // deserialize("mnist_network.dat") >> net;
+
+
+    // Now let's run the training images through the network.  This statement runs all the
+    // images through it and asks the loss layer to convert the network's raw output into
+    // labels.  In our case, these labels are the numbers between 0 and 9.
+    std::vector<unsigned long> predicted_labels = net(training_images);
+    int num_right = 0;
+    int num_wrong = 0;
+    // And then let's see if it classified them correctly.
+    for (size_t i = 0; i < training_images.size(); ++i)
+    {
+        if (predicted_labels[i] == training_labels[i])
+            ++num_right;
+        else
+            ++num_wrong;
+        
+    }
+    cout << "training num_right: " << num_right << endl;
+    cout << "training num_wrong: " << num_wrong << endl;
+    cout << "training accuracy:  " << num_right/(double)(num_right+num_wrong) << endl;
+
+    // Let's also see if the network can correctly classify the testing images.  Since
+    // MNIST is an easy dataset, we should see at least 99% accuracy.
+    predicted_labels = net(testing_images);
+    num_right = 0;
+    num_wrong = 0;
+    for (size_t i = 0; i < testing_images.size(); ++i)
+    {
+        if (predicted_labels[i] == testing_labels[i])
+            ++num_right;
+        else
+            ++num_wrong;
+        
+    }
+    cout << "testing num_right: " << num_right << endl;
+    cout << "testing num_wrong: " << num_wrong << endl;
+    cout << "testing accuracy:  " << num_right/(double)(num_right+num_wrong) << endl;
+
+}
+catch(std::exception& e)
+{
+    cout << e.what() << endl;
+}
+