Update inctructios

.github/workflows/ci.yml

@@ -29,6 +29,10 @@ jobs:
     - uses: actions/checkout@v4
       with:
         submodules: true
+    - name: Set binary paths
+      id: set_binaries
+      run: |
+        echo "ACC_BINARY=build/bin/ACC" >> $GITHUB_OUTPUT
     - name: Setup ccache
       uses: hendrikmuhs/ccache-action@v1.2
       with:
@@ -59,7 +63,7 @@ jobs:
       with:
        name: mnist-${{ matrix.build_type }}${{ matrix.stats && '-stats' || '' }}
        path: |
-        build/bin/ACC_MNIST*
+        ${{ steps.set_binaries.outputs.ACC_BINARY }}
         build/bin/opencv_libs/*
         build/setenv.sh
     - name: Test
@@ -227,7 +231,10 @@ jobs:
     - uses: actions/checkout@v4
       with:
         fetch-depth: 0
-
+    - name: Set binary path
+      id: set_eval_binary
+      run: |
+        echo "EVAL_BINARY=build/bin/ACC" >> $GITHUB_OUTPUT
     - name: Install system dependencies
       run: |
         sudo apt-get update
@@ -274,7 +281,7 @@ jobs:
 
     - name: Prepare environment
       run: |
-        chmod +x build/bin/ACC_MNIST*
+        chmod +x "${{ steps.set_eval_binary.outputs.EVAL_BINARY }}"
 
         export LD_LIBRARY_PATH=$PWD/build/bin/opencv_libs:/usr/lib/x86_64-linux-gnu
         echo "Final LD_LIBRARY_PATH: $LD_LIBRARY_PATH"
@@ -290,12 +297,12 @@ jobs:
        export LD_LIBRARY_PATH=$PWD/build/bin/opencv_libs:/usr/lib/x86_64-linux-gnu
        echo "LD_LIBRARY_PATH: $LD_LIBRARY_PATH"
 
-       LD_DEBUG=files ./build/bin/ACC_MNIST* 2> ld_debug.log
+       LD_DEBUG=files "${{ steps.set_eval_binary.outputs.EVAL_BINARY }}" --model alexnet_mnist 2> ld_debug.log
        echo "### Library loading debug ###"
        grep -i "opencv_imgcodecs" ld_debug.log
 
-       ./build/bin/ACC_MNIST* > accuracy.txt
-       echo "Accuracy: $(cat accuracy.txt)"
+       "${{ steps.set_eval_binary.outputs.EVAL_BINARY }}" > accuracy.txt
+        echo "Accuracy: $(cat accuracy.txt)"
 
     - name: Update README (master only)
       if: github.ref == 'refs/heads/master'

README.md

@@ -6,16 +6,18 @@
 
 <!--ACCURACY_PLACEHOLDER-->Accuracy: Stat: 98.01% (updated: 2025-04-28)<!--END_ACCURACY-->
 ## Short description
-A lightweight C++ library for performing high-performance inference on MNIST handwritten digits using a modified AlexNet architecture. Designed for efficiency and educational purposes, this project demonstrates how classic CNNs can be optimized for small-scale tasks in native environments.
+A lightweight C++ library for performing high-performance inference on MNIST and ImageNet using a modified AlexNet, different ONNX and Yolo architectures. Designed for efficiency and educational purposes, this project demonstrates how classic CNNs can be optimized for small-scale tasks in native environments.
 ### Key Features:
 
 * C++17 implementation for bare-metal performance
 
 * Simplified AlexNet for 28×28 grayscale images
 
+* Googlenet, Densenet, Resnet and Yolo11x-cls for images of any size
+
 * Parallel computing via Intel OneTBB (Threading Building Blocks)
 
-* Pre-trained model: AlexNet-model.h5 included
+* Pre-trained model: AlexNet-model.h5, Googlenet included
 ## **Some files used to create the library**
 ### Neural network models
 You need to download [Alexnet-model.h5](https://github.com/moizahmed97/Convolutional-Neural-Net-Designer/blob/master/AlexNet-model.h5) to the folder *docs*
@@ -30,9 +32,9 @@ Other models:</br>
 
 ## **How do I launch the inference?**
 * Make sure you install the project dependencies by running: *pip install -r requirements.txt*
-* You need to run the script *parser.py* that is located in app/AlexNet to read weights from a model *Alexnet-model.h5* and the json file with the weights will be stored in the *docs* folder.
+* You need to run the script *parser.py* that is located in app/converters to read weights from a model *Alexnet-model.h5* or *parser_onnx.py* to read weights from a models ONNX or YOLO and the json file with the weights will be stored in the *docs* folder.
 * Then put the test images in png format in the folder *docs/input*
-* After building the project, which is described below, run Graph_build in folder *build/bin*
+* After building the project, which is described below, run Graph_build with the parameter --model (alexnet_mnist or googlenet or densenet or resnet or yolo) and the parameter --parallel if you need. App Graph_build is located in folder *build/bin*
 
 ## **Building a Project**
 ### *Windows*
@@ -69,7 +71,7 @@ To build and run this project locally on Windows, follow these steps:
     ```
    and run the file
     ```bash
-   Graph_Build.exe
+   Graph_Build.exe --model alexnet_mnist
     ```
 ### *Linux/macOS*
    To build and run this project locally on Linux or macOS, follow these steps:
@@ -116,7 +118,7 @@ To build and run this project locally on Windows, follow these steps:
     ```
    and run the file
     ```bash
-   ./Graph_Build
+   ./Graph_Build --model alexnet_mnist
     ```
 
 ## Test Process
@@ -147,10 +149,14 @@ To start the testing process locally, you need to go to the directory
    ./run_test
    ```
 
-## **Accuracy validation**
+## **Accuracy validation for Alexnet on MNIST**
 To run accuracy validation you need to use the MNIST dataset, which you can download [here](https://github.com/DeepTrackAI/MNIST_dataset/tree/main/mnist/test) and put it in a folder *docs/mnist/mnist/test*
-Now you can run accuracy check - *build\bin\ACC_MNIST.exe*
-* **The accuracy should be 98.02%**
+Now you can run accuracy check - *build\bin\ACC.exe --model alexnet_mnist*
+* **The accuracy should be 98.01%**
+
+## **Accuracy validation for ONNX or YOLO models on ImageNet**
+To run accuracy validation you need to use the ImageNet dataset, which you can download [here](https://www.kaggle.com/datasets/sautkin/imagenet1kvalid) and put it in a folder *docs/Imagenet/*
+Now you can run accuracy check - *build\bin\ACC.exe --model googlenet*
 
 ## **Documentation of project**
 https://github.com/embedded-dev-research/ITLabAI/blob/Semyon1104/Final_documentation/docs/IT_Lab_2023.pdf

app/Converters/parser_onnx.py

@@ -5,15 +5,17 @@
 from onnx import helper, numpy_helper
 from ultralytics import YOLO
 
+
 def convert_pt_to_onnx(pt_model_path, onnx_model_path=None):
     if onnx_model_path is None:
         onnx_model_path = pt_model_path.replace('.pt', '.onnx')
 
     model = YOLO(pt_model_path)
-    model.export(format="onnx", dynamic=False, simplify=True)
+    model.export(format="onnx", dynamic=False, simplify=False)
 
     return onnx_model_path
 
+
 def onnx_to_json(model_path, output_json_path):
     if model_path.endswith('.pt'):
         model_path = convert_pt_to_onnx(model_path)
@@ -31,12 +33,40 @@ def onnx_to_json(model_path, output_json_path):
     }
 
     layer_info = []
+
+    input_info = {}
+    for input in model.graph.input:
+        if input.name in initializers_dict:
+            continue
+
+        shape = []
+        for dim in input.type.tensor_type.shape.dim:
+            if dim.HasField('dim_value'):
+                # 0 означает динамическую размерность в ONNX
+                shape.append(dim.dim_value if dim.dim_value != 0 else -1)
+            elif dim.HasField('dim_param'):
+                # Обрабатываем именованные параметры размерностей
+                shape.append(-1)  # или можно сохранить как строку: dim.dim_param
+            else:
+                shape.append(-1)  # неизвестная размерность
+
+        input_info = {
+            "name": input.name,
+            "shape": shape,
+            "data_type": input.type.tensor_type.elem_type
+        }
+        break
+
     input_layer = {
         "index": 0,
-        "name": "input_1",
+        "name": input_info.get("name", "input_1"),
         "type": "InputLayer",
         "weights": [],
-        "attributes": {}
+        "bias": [],
+        "attributes": {
+            "shape": input_info.get("shape", []),
+            "data_type": input_info.get("data_type", 1)
+        }
     }
     layer_info.append(input_layer)
 
@@ -45,9 +75,14 @@ def onnx_to_json(model_path, output_json_path):
             "index": len(layer_info),
             "name": node.name.replace('/', '_'),
             "type": node.op_type,
-            "attributes": {}
+            "attributes": {},
+            "inputs": []
         }
 
+        for input_name in node.input:
+            if input_name not in initializers_dict:
+                layer_data["inputs"].append(input_name.replace('/', '_'))
+
         for attr in node.attribute:
             attr_value = helper.get_attribute_value(attr)
             if isinstance(attr_value, TensorProto):
@@ -67,29 +102,44 @@ def onnx_to_json(model_path, output_json_path):
             elif attr.name == "strides":
                 layer_data["strides"] = attr_value
 
-        node_init = []
-        for input_name in node.input:
-            if input_name in initializers_dict:
-                node_init.append(initializers_dict[input_name])
-
-        if len(node_init) == 1:
-            init = node_init[0]
-            if len(init["dims"]) == 0 or (len(init["dims"]) == 1 and init["dims"][0] == 1):
-                layer_data["value"] = init["values"] if len(init["dims"]) == 0 else init["values"][0]
-            else:
-                layer_data["weights"] = init["values"]
-        elif len(node_init) > 1:
-            weights = []
-            for init in node_init[:-1]:
-                if len(init["dims"]) > 0:
-                    weights.extend(init["values"]) if isinstance(init["values"][0], list) else weights.append(
-                        init["values"])
-
-            if weights:
-                layer_data["weights"] = weights
-
-            if len(node_init[-1]["dims"]) == 1:
-                layer_data["bias"] = node_init[-1]["values"]
+        if node.op_type == "BatchNormalization":
+            bn_params = []
+            for input_name in node.input:
+                if input_name in initializers_dict:
+                    bn_params.append(initializers_dict[input_name])
+
+            if len(bn_params) >= 4:
+                layer_data["scale"] = bn_params[0]["values"]
+                layer_data["bias"] = bn_params[1]["values"]
+                layer_data["mean"] = bn_params[2]["values"]
+                layer_data["var"] = bn_params[3]["values"]
+
+                layer_data["weights"] = []
+
+        else:
+            node_init = []
+            for input_name in node.input:
+                if input_name in initializers_dict:
+                    node_init.append(initializers_dict[input_name])
+
+            if len(node_init) == 1:
+                init = node_init[0]
+                if len(init["dims"]) == 0 or (len(init["dims"]) == 1 and init["dims"][0] == 1):
+                    layer_data["value"] = init["values"] if len(init["dims"]) == 0 else init["values"][0]
+                else:
+                    layer_data["weights"] = init["values"]
+            elif len(node_init) > 1:
+                weights = []
+                for init in node_init[:-1]:
+                    if len(init["dims"]) > 0:
+                        weights.extend(init["values"]) if isinstance(init["values"][0], list) else weights.append(
+                            init["values"])
+
+                if weights:
+                    layer_data["weights"] = weights
+
+                if len(node_init[-1]["dims"]) == 1:
+                    layer_data["bias"] = node_init[-1]["values"]
 
         layer_info.append(layer_data)
 
@@ -116,7 +166,7 @@ def default(self, obj):
 
 BASE_DIR = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
 
-MODEL_PATH = os.path.join(BASE_DIR, 'docs\\models', 'yolo11x-cls.pt')
-MODEL_DATA_PATH = os.path.join(BASE_DIR, 'docs\\jsons', 'yolo11x-cls_onnx_model.json')
+MODEL_PATH = os.path.join(BASE_DIR, 'docs\\models', 'resnest101e_Opset16.onnx')
+MODEL_DATA_PATH = os.path.join(BASE_DIR, 'docs\\jsons', 'resnest101e_Opset16_onnx_model.json')
 
 onnx_to_json(MODEL_PATH, MODEL_DATA_PATH)

app/Graph/CMakeLists.txt

@@ -18,8 +18,8 @@ target_include_directories(BuildGraph PUBLIC ${CMAKE_SOURCE_DIR}/3rdparty/Json/i
 add_executable(Graph_Build graph_build.cpp)
 target_link_libraries(Graph_Build BuildGraph)
 
-add_executable(ACC_MNIST acc_check_mnist.cpp)
-target_link_libraries(ACC_MNIST BuildGraph)
+add_executable(ACC acc_check.cpp)
+target_link_libraries(ACC BuildGraph)
 
 if (NOT CMAKE_BUILD_TYPE)
     set(CMAKE_BUILD_TYPE "Debug")
@@ -33,24 +33,41 @@ if (WIN32)
 endif()
 
 if (WIN32)
-    add_custom_command(TARGET ACC_MNIST POST_BUILD
+	if ("${CMAKE_BUILD_TYPE}" STREQUAL "DEBUG")
+	    set(CMAKE_BUILD_TYPE "Debug")
+	endif()
+    if ("${CMAKE_BUILD_TYPE}" STREQUAL "RELEASE")
+	    set(CMAKE_BUILD_TYPE "Release")
+	endif()
+    add_custom_command(TARGET ACC POST_BUILD
         COMMAND ${CMAKE_COMMAND} -E copy_directory
             "${OPENCV_BUILD_DIR}/bin/."
             "${CMAKE_BINARY_DIR}/bin/")
 endif()
 
 file(DOWNLOAD
     "https://raw.githubusercontent.com/DeepTrackAI/MNIST_dataset/main/mnist/test/1_000008.png"
-    "${CMAKE_SOURCE_DIR}/docs/input/test1.png"
+    "${CMAKE_SOURCE_DIR}/docs/input/28/test1.png"
     SHOW_PROGRESS
     STATUS status_code
     LOG log_file
 )
 
-add_definitions(-DIMAGE1_PATH="${CMAKE_SOURCE_DIR}/docs/input/")
+file(DOWNLOAD
+    "blob:https://ru.pinterest.com/63b88674-b4a6-4ef3-85b2-ab57ef7bb8e7"
+    "${CMAKE_SOURCE_DIR}/docs/input/Imagenet_test/tench.png"
+    SHOW_PROGRESS
+    STATUS status_code
+    LOG log_file
+)
+
+add_definitions(-DIMAGE28_PATH="${CMAKE_SOURCE_DIR}/docs/input/28/")
+add_definitions(-DIMAGENET_ACC="${CMAKE_SOURCE_DIR}/docs/ImageNet/test/")
+add_definitions(-DIMAGENET_PATH="${CMAKE_SOURCE_DIR}/docs/input/Imagenet_test/")
 add_definitions(-DMODEL_PATH_H5="${CMAKE_SOURCE_DIR}/docs/jsons/model_data_alexnet_1.json")
 add_definitions(-DMODEL_PATH_GOOGLENET_ONNX="${CMAKE_SOURCE_DIR}/docs/jsons/googlenet_onnx_model.json")
 add_definitions(-DMODEL_PATH_DENSENET_ONNX="${CMAKE_SOURCE_DIR}/docs/jsons/densenet121_Opset16_onnx_model.json")
 add_definitions(-DMODEL_PATH_RESNET_ONNX="${CMAKE_SOURCE_DIR}/docs/jsons/resnest101e_Opset16_onnx_model.json")
 add_definitions(-DMODEL_PATH_YOLO11NET_ONNX="${CMAKE_SOURCE_DIR}/docs/jsons/yolo11x-cls_onnx_model.json")
+add_definitions(-DIMAGENET_LABELS="${CMAKE_SOURCE_DIR}/docs/imagenet1000_clsidx_to_labels.json")
 add_definitions(-DMNIST_PATH="${CMAKE_SOURCE_DIR}/docs/mnist/mnist/test")