### Setup for TIDL Tools

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/README.md

Details the prerequisites and setup steps required before building and running the TIDL C++ examples. This includes installing the libyaml-cpp-dev library and preparing the environment by following instructions in the main README.

```bash
sudo apt-get install libyaml-cpp-dev
```

--------------------------------

### Advanced Setup: Skipping C++ Dependencies

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/docs/advanced_setup.md

Invokes the setup script with an option to validate only Python examples and skip the execution of C++ examples.

```bash
source ./setup.sh --skip_cpp_deps
```

--------------------------------

### Advanced Setup Options

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/README.md

Details setup options for advanced users aiming to optimize their setup time. This guide is for users who need to fine-tune their environment.

```markdown
[**Advanced Setup Options**](docs/advanced_setup.md#advanced-setup-options) - Setup options for advanced users to optimize setup time
```

--------------------------------

### Setup TIDL Tools Environment

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/ti_edgeai_tidl_tools_am62a.ipynb

Sets the System-on-Chip (SOC) to 'am62a' and sources the setup script for the TIDL tools. It skips downloading ARM GCC and C++ dependencies.

```shell
!export SOC=am62a && cd edgeai-tidl-tools && source ./setup.sh  --skip_arm_gcc_download --skip_cpp_deps
```

--------------------------------

### Build CPP Examples with CMake

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/README.md

Instructions for building C++ examples using CMake. It details creating a build directory, using CMake flags for different compilation modes and target devices, and specifying installation paths for dependencies.

```bash
mkdir build && cd build
cmake -DFLAG1=val -DFLAG2=val ../examples
make -j2
cd ../
```

```bash
cmake -DTARGET_DEVICE=am62 TARGET_CPU=arm ../examples
make -j2
cd ../
```

```bash
cmake -DTARGET_DEVICE=am62  -DARMNN_ENABLE=1../examples    
```

```bash
cmake -DTARGET_DEVICE=am62 TARGET_CPU=arm -DARMNN_ENABLE=1../examples
```

```bash
cd <target_fs>/usr/lib
ln -s libonnxruntime.so.1.7.0 libonnxruntime.so
```

--------------------------------

### Setup EdgeAI TIDL Tools Environment

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/ti_edgeai_tidl_tools_am68a.ipynb

Sets the SOC environment variable and sources the setup script for the EdgeAI TIDL Tools. It also skips downloading ARM GCC and C++ dependencies.

```bash
!export SOC=am68a && cd edgeai-tidl-tools && source ./setup.sh  --skip_arm_gcc_download --skip_cpp_deps
```

--------------------------------

### Setup EdgeAI TIDL Tools Environment

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/ti_edgeai_tidl_tools_am69a.ipynb

Sets the System-on-Chip (SOC) to 'am69a' and sources the setup script for the EdgeAI TIDL Tools. It skips downloading ARM GCC and C++ dependencies.

```shell
!export SOC=am69a && cd edgeai-tidl-tools && source ./setup.sh  --skip_arm_gcc_download --skip_cpp_deps
```

--------------------------------

### Download Model and Data, Setup Environment

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/infer_tflite.ipynb

Downloads the EfficientNet-Lite0 TFLite model and a sample image (jet.jpeg). It also renames the downloaded model file and installs the required tflite_runtime library.

```python
!wget https://akm-img-a-in.tosshub.com/indiatoday/images/story/201804/jet.jpeg
!wget https://tfhub.dev/tensorflow/lite-model/efficientnet/lite0/fp32/2?lite-format=tflite
!mv /content/2?lite-format=tflite efficientnet-lit0.tflite
!pip3 install tflite_runtime==2.8.0
```

--------------------------------

### Setup TIDL ONNX Model Optimizer

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/osrt-model-tools/osrt_model_tools/onnx_tools/tidl_onnx_model_optimizer/README.md

Sets up the TIDL ONNX Model Optimizer by navigating to the 'onnx_tools' directory and sourcing the 'setup.sh' script. This installs the 'tidl_onnx_model_optimizer' Python package.

```shell
cd ../onnx_tools
source ./setup.sh
```

--------------------------------

### Clone EdgeAI TIDL Tools Repository

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/ti_edgeai_tidl_tools_am68a.ipynb

Clones the specified branch of the EdgeAI TIDL Tools repository from GitHub. This is the initial step to get the necessary tools and examples.

```bash
!git clone --single-branch --depth 1 --branch master https://github.com/TexasInstruments/edgeai-tidl-tools.git
```

--------------------------------

### Install Dependencies and Download Model/Data

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/infer_ort.ipynb

Installs the onnxruntime library and downloads the SqueezeNet ONNX model and a sample image (jet.jpeg). It also renames the downloaded model file.

```python
!pip install onnxruntime
!wget https://akm-img-a-in.tosshub.com/indiatoday/images/story/201804/jet.jpeg
!wget https://github.com/onnx/models/blob/1008b6e9450ab632538390e38e0ab28b3f207fbc/vision/classification/squeezenet/model/squeezenet1.1-7.onnx?raw=true
!mv squeezenet1.1-7.onnx?raw=true squeezenet1.1-7.onnx
```

--------------------------------

### Setup TIDL ONNX Model Utilities

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/osrt-model-tools/osrt_model_tools/onnx_tools/tidl_onnx_model_utils/README.md

Installs the tidl_onnx_model_utils Python package by navigating to the onnx_tools directory and sourcing the setup script. This process depends on the onnx-graphsurgeon library.

```bash
cd ../onnx_tools
source ./setup.sh
```

--------------------------------

### Docker One-Time Setup Script

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/docs/advanced_setup.md

Executes the one-time setup script for Docker, which prepares the environment for building and running Docker images.

```bash
source ./scripts/docker/setup_docker.sh
```

--------------------------------

### Post-Container Run Setup

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/docs/advanced_setup.md

Configures the environment within the running Docker container by exporting SOC and TIDL_TOOLS_TYPE, and then sourcing the setup script. This needs to be done every time the container is re-run.

```bash
export SOC=<Your SOC name>
export TIDL_TOOLS_TYPE=<Your Tools Type (CPU|GPU)>
source ./setup.sh
```

--------------------------------

### Setup TIDL Tools Environment for AM68PA

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/ti_edgeai_tidl_tools_tda4vm.ipynb

Sets the System-on-Chip (SOC) to 'am68pa' and sources the setup script for the TIDL tools. This configures the necessary environment variables and dependencies for running TIDL on the specified hardware, skipping ARM GCC download and C++ dependencies.

```shell
!export SOC=am68pa && cd edgeai-tidl-tools && source ./setup.sh  --skip_arm_gcc_download --skip_cpp_deps
```

--------------------------------

### Build TIDLRT CPP Examples

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/tidlrt_cpp/advanced_examples/README.md

Builds the C++ examples for TIDLRT using CMake. This process involves creating a build directory, configuring the build with CMake, and then compiling the project using make.

```bash
mkdir build && cd build
cmake ../examples/
make -j2 
cd  ../
```

--------------------------------

### Advanced Setup: Using Existing ARM GCC Toolchain

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/docs/advanced_setup.md

Configures the setup script to use an existing ARM GCC toolchain by setting the ARM64_GCC_PATH environment variable and using the --skip_arm_gcc_download option.

```bash
export ARM64_GCC_PATH=$(pwd)/arm-gnu-toolchain-13.2.Rel1-x86_64-aarch64-none-linux-gnu
source ./setup.sh --skip_arm_gcc_download
```

--------------------------------

### Build CPP Advanced Examples

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/advanced_examples/README.md

Builds the C++ advanced examples using CMake. This involves creating a build directory, configuring the build with CMake, and then compiling the project.

```bash
mkdir build && cd build
cmake ../examples/
make
cd  ../
```

--------------------------------

### Run TFLite Delegate Example (with compilation)

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/ti_edgeai_tidl_tools_am68a.ipynb

Executes a TFLite delegate example using Python. It sets environment variables for the SOC, library path, and TIDL tools path, then runs the `tflrt_delegate.py` script with a specific model, including compilation.

```bash
!export SOC=am68a && \
export LD_LIBRARY_PATH=/content/edgeai-tidl-tools/tidl_tools && \
export TIDL_TOOLS_PATH=/content/edgeai-tidl-tools/tidl_tools && \
cd ./edgeai-tidl-tools/examples/osrt_python/tfl && \
python3 tflrt_delegate.py -c -m cl-tfl-mobilenet_v1_1.0_224
```

--------------------------------

### Build TIDLRT CPP Examples

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/tidlrt_cpp/README.md

Instructions to build the C++ examples for TIDLRT using CMake. This involves creating a build directory, configuring the build with CMake, and then compiling the project with make.

```bash
mkdir build && cd build
cmake ../examples/
make
cd  ../
```

--------------------------------

### Setup Environment

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/osrt-model-tools/osrt_model_tools/onnx_tools/README.md

Executes the setup script to configure the environment for using the ONNX tools. This is a prerequisite for running the optimization and utility packages.

```bash
source ./setup.sh
```

--------------------------------

### Run TIDL CPP Inference Examples

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/README.md

Commands to execute pre-compiled C++ inference examples for various models (ResNet18, MobileNet, InceptionNet, SSD-Lite, DeepLabV3) using different runtimes (ORT, TFL, DLR).

```bash
./bin/Release/ort_main -f model-artifacts/cl-ort-resnet18-v1/artifacts  -i test_data/airshow.jpg
```

```bash
./bin/Release/tfl_main -f model-artifacts/cl-tfl-mobilenet_v1_1.0_224/artifacts -i test_data/airshow.jpg
```

```bash
./bin/Release/dlr_main -f model-artifacts/cl-dlr-tflite_inceptionnetv3/artifacts  -i test_data/airshow.jpg
```

```bash
./bin/Release/dlr_main -f model-artifacts/cl-dlr-onnx_mobilenetv2/artifacts  -i test_data/airshow.jpg
```

```bash
./bin/Release/ort_main -f model-artifacts/od-ort-ssd-lite_mobilenetv2_fpn/artifacts -i test_data/ADE_val_00001801.jpg
```

```bash
./bin/Release/tfl_main -f model-artifacts/od-tfl-ssd_mobilenet_v2_300_float/artifacts -i test_data/ADE_val_00001801.jpg
```

```bash
./bin/Release/tfl_main -f model-artifacts/ss-tfl-deeplabv3_mnv2_ade20k_float/artifacts -i test_data/ADE_val_00001801.jpg
```

--------------------------------

### Run CPP Advanced Examples

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/advanced_examples/README.md

Executes the C++ advanced applications. The examples demonstrate priority scheduling for multiple models, with options to specify model paths, priorities, threads, offloading to a device, and preempt delays.

```bash
./bin/Release/tfl_priority_scheduling -h
./bin/Release/ort_main_adv -h
```

```bash
./bin/Release/tfl_priority_scheduling -i test_data/ADE_val_00001801.jpg test_data/airshow.jpg -m  model-artifacts/ss-tfl-deeplabv3_mnv2_ade20k_float/artifacts/ model-artifacts/cl-tfl-mobilenet_v1_1.0_224/artifacts/ -p 1 0 -t 1  -a 1  -d 1 -e 3
```

--------------------------------

### Run TFLite Delegate Example (with compilation)

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/ti_edgeai_tidl_tools_am62a.ipynb

Executes a TFLite delegate example using Python. It sets environment variables for library paths and the TIDL tools path, then runs the 'tflrt_delegate.py' script with a specific model ('cl-tfl-mobilenet_v1_1.0_224') and compilation flag.

```shell
!export SOC=am62a && \
export LD_LIBRARY_PATH=/content/edgeai-tidl-tools/tidl_tools && \
export TIDL_TOOLS_PATH=/content/edgeai-tidl-tools/tidl_tools && \
cd ./edgeai-tidl-tools/examples/osrt_python/tfl && \
python3 tflrt_delegate.py -c -m cl-tfl-mobilenet_v1_1.0_224
```

--------------------------------

### Cloning and Checking Out Repository

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/docs/advanced_setup.md

Steps to clone the edgeai-tidl-tools GitHub repository and checkout a specific tag compatible with the SDK version.

```bash
git clone https://github.com/TexasInstruments/edgeai-tidl-tools.git
cd edgeai-tidl-tools
git checkout <TAG Compatible with your SDK version>
```

--------------------------------

### Running the Docker Image

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/docs/advanced_setup.md

Launches the Docker container for the edgeai-tidl-tools project.

```bash
source ./scripts/docker/run_docker.sh
```

--------------------------------

### Troubleshooting GPU Docker Setup: Restarting Docker

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/docs/advanced_setup.md

Restarts the Docker service after installing the nvidia-container-toolkit to ensure GPU detection works correctly.

```bash
sudo systemctl restart docker
```

--------------------------------

### Install Requirements

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_python/advanced_examples/unit_tests_validation/README.md

Installs necessary Python packages for the unit test validation scripts using a requirements file.

```bash
cd examples/osrt_python/advanced_examples/unit_tests_validation
pip3 install -r ./requirements.txt
```

--------------------------------

### Prepare Configuration Files

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/tidlrt_tools.ipynb

Appends configuration settings to a config file and creates a list of input data for quantization/calibration. The configuration specifies the path to the device configuration file and the input data list points to the image and its expected output index.

```python
!echo "perfSimConfig = /content/tidl_tools/device_config.cfg" >> /content/resnet18/config
!echo "/content/resnet18/jet.jpeg 895" >> /content/resnet18/in_data_list.txt
```

--------------------------------

### Example Quantization Proto File

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/docs/tidl_quantParams.md

An example of a valid prototxt file for a dummy model with two layers, demonstrating the structure and required fields for bypassing TIDL's calibration step.

```text
num_layers: 2
quant_type: SYMMETRIC
calib_type: PERTENSOR
layers {
  layer_name: "data"
  layer_type: "TIDL_DataLayer"
  bit_depth: 1
  outputs {
    min: 0
    max: 255
    size: 1
    element_type: 0
    scale: 1
    zero_point: 0
  }
}
layers {
  layer_name: "conv1a"
  layer_type: "TIDL_ConvolutionLayer"
  bit_depth: 1
  outputs {
    min: 0
    max: 4.09663105
    size: 1
    element_type: 0
    scale: 56.387794494628906
    zero_point: 0
  }
  weights {
    min: -0.00830631796
    max: 0.00886716694
    size: 1
    element_type: 0
    scale: 14435.275390625
    zero_point: 0
  }
  bias {
    min: -0.52766174077987671
    max: 0.53365200757980347
    size: 3
    element_type: 0
    scale: 120.27982330322266
    zero_point: 0
    value: 0.28929620981216431
    value: -0.52766174077987671
    value: 0.53365200757980347
  }
}

```

--------------------------------

### Deploy Model Artifacts to EVM

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/README.md

This snippet outlines the necessary folders to copy from the host PC to the EVM (Embedded Linux Development Board) for running the model examples on the target hardware.

```shell
./model-artifacts
./models
```

--------------------------------

### Building the Docker Image

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/docs/advanced_setup.md

Builds the Docker image for the edgeai-tidl-tools project. Requires setting the repository location and proxy environment variables before execution.

```bash
export REPO_LOCATION=artifactory.itg.ti.com/docker-public/library/
export PROXY=http://webproxy.ext.ti.com:80
source ./scripts/docker/build_docker.sh
```

--------------------------------

### Build utils_adv Library and Install

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/advanced_examples/utils/CMakeLists.txt

This snippet defines the build process for the 'utils_adv' library, which includes source files for argument parsing and print utilities. It also specifies the installation targets for the library and its header files.

```cmake
cmake_minimum_required(VERSION 3.0.2)
include(${CMAKE_CURRENT_SOURCE_DIR}/../../../cmake/common.cmake)
add_library(utils_adv ${CMAKE_CURRENT_SOURCE_DIR}/src/arg_parsing.cpp
                            ${CMAKE_CURRENT_SOURCE_DIR}/include/arg_parsing.h
                            ${CMAKE_CURRENT_SOURCE_DIR}/src/print_utils.cpp
                            ${CMAKE_CURRENT_SOURCE_DIR}/include/print_utils.h
                            )

install(TARGETS
         utils_adv
         DESTINATION lib)

install(FILES 
${CMAKE_CURRENT_SOURCE_DIR}/include/arg_parsing.h
${CMAKE_CURRENT_SOURCE_DIR}/include/print_utils.h

DESTINATION include)
```

--------------------------------

### TIDL Required Options

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_python/README.md

These options must be specified by the user when creating inference sessions for respective runtimes. They are crucial for model compilation and inference setup.

```APIDOC
tidl_tools_path:
  Description: Path to TIDL tools for model compilation. Populated using ${TIDL_TOOLS_PATH} environment variable.
  Option Type: Model compilation
  Default: Not specified (must be provided by user)

artifacts_folder:
  Description: Path to the folder where model artifacts are saved or to be saved.
  Option Type: Model compilation / Model inference
  Default: Not specified (must be provided by user)
```

--------------------------------

### Python API and Examples

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/README.md

Provides detailed documentation for all compile and inference options for TIDL offload for each runtime session using Python. This guide is essential for users working with Python-based workflows.

```markdown
[**Python API and examples**](examples/osrt_python/README.md) - Detailed documentation on all the compile and inference options for TIDL offload for each runtime sessions
```

--------------------------------

### Initial Setup and Environment Configuration

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/README.md

Clones the TIDL tools repository, checks out a compatible tag, sets the SOC environment variable, and sources the setup script to configure the environment. The 'source' command is crucial for exporting environment variables.

```bash
git clone https://github.com/TexasInstruments/edgeai-tidl-tools.git
cd edgeai-tidl-tools
git checkout <TAG Compatible with your SDK version>
# Supported SOC name strings am62, am62a, (am68a or j721s2), (am68pa or j721e), (am69a or j784s4), (am67a or j722s)
export SOC=<Your SOC name>
source ./setup.sh
```

--------------------------------

### TIDL Multi-DSP Inference Options

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_python/README.md

Configuration options for devices with multiple DSP cores, affecting inference mode, core utilization, and execution start points. These options are crucial for optimizing performance on multi-core architectures.

```APIDOC
advanced_options:inference_mode
  Description: Specifies the feature/mode for inference. Must be set during compilation and affects generated artifacts.
  Supported values:
    0 (TIDL_inferenceModeDefault): Default inference mode.
    1 (TIDL_inferenceModeHighThroughput): High throughput inference mode.
    2 (TIDL_inferenceModeLowLatency): Low latency inference mode.
  Default value: 0
  Option Type: Model compilation
  Additional details: Refer [Multi-DSP inference](../../docs/tidl_fsg_multi_c7x.md) for more details.

advanced_options:num_cores
  Description: Specifies the number of DSP cores to be used for inference.
  Supported values/range: Min: 1, Max: maximum number of DSP cores available on the device.
  Default value: 1
  Option Type: Model compilation
  Additional details: Refer [Multi-DSP inference](../../docs/tidl_fsg_multi_c7x.md) for more details.

core_number
  Description: Specifies the index of the core out of all available C7x cores to be used for single-core inference. This option is 1-indexed. For example, core_number = 1 for C7x_1.
  Supported values/range: Min: 1, Max: maximum number of DSP cores available on the device.
  Default value: 1
  Option Type: Model inference
  Additional details: Refer [Multi-DSP inference](../../docs/tidl_fsg_multi_c7x.md) for more details.

core_start_idx
  Description: Specifies the index of the core from which to start processing. In case of inference_mode = 1 or 2, execution would happen on C7x_{core_start_idx} to C7x_{core_start_idx + advanced_options:num_cores}.
  Supported values/range: Min: 1, Max: maximum number of DSP cores available on the device.
  Default value: 1
  Option Type: Model inference
  Additional details: Refer [Multi-DSP inference](../../docs/tidl_fsg_multi_c7x.md) for more details.
```

--------------------------------

### Basic CMake Project Setup

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/dlr/CMakeLists.txt

This snippet demonstrates the fundamental CMake commands to initialize a project. It sets the minimum required CMake version, defines the project name, and includes common build scripts from a specified directory. It then defines the source files for the application and initiates the build process.

```cmake
cmake_minimum_required(VERSION 3.0.2)

set(PROJ_NAME dlr_main)

project(${PROJ_NAME})

include(${CMAKE_CURRENT_SOURCE_DIR}/../../cmake/common.cmake)

set(DLR_CLS_APP_SRCS
    dlr_main.cpp)

build_app(${PROJ_NAME} DLR_CLS_APP_SRCS)
```

--------------------------------

### Enable ARM Only Mode for Model Compilation

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/docs/tidl_osr_debug.md

This snippet demonstrates how to enable ARM-only mode for model compilation, which is useful for verifying script functionality without C7x offload. It shows the command for out-of-the-box ONNX examples and advises referring to custom model documentation for enabling this mode.

```python
python3 onnxrt_ep.py -d
```

--------------------------------

### Get All Node Names from ONNX Model

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/osrt-model-tools/osrt_model_tools/onnx_tools/tidl_onnx_model_utils/README.md

Retrieves a comma-separated string of all node names within an ONNX model, between specified start and end layers (inclusive). If the end node is 'None', model output nodes are assumed. Handles cases where specified end nodes are not found.

```python
from tidl_onnx_model_utils import get_all_node_names

# Example usage:
model_path = "path/to/your/model.onnx"
start_end_layers = {
    '/bbox_head/Sigmoid' : ['/Cast_3', '/GatherElements_1'],
    '/bbox_head/offset_head/offset_head.0/Conv' : None,
    '/Div_2' : ['/Mul_5']
}

deny_list = get_all_node_names(model_path, start_end_layers)
print(deny_list)
```

--------------------------------

### Deploy and Run Inference on EVM Board

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/jupyter_notebooks/colab/tidlrt_tools.ipynb

Provides instructions for deploying the generated model artifacts to an EVM board and running inference. It includes commands to navigate to the vision_apps directory, source initialization scripts, change directory, set the LD_LIBRARY_PATH, prepare input data, and execute the inference using the device-specific algorithm test tool.

```bash
root@ j7-evm:~# cd /opt/vision_apps
root@ j7-evm:~# source ./vision_apps_init.sh
root@ j7-evm:~# cd /opt/tidl_test
root@ j7-evm:/opt/tidl_test# export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib
root@ j7-evm:~# echo "/opt/resnet18/jet.jpeg 895" > /opt/resnet18/in_data_list.txt
root@ j7-evm:/opt/tidl_test# ./TI_DEVICE_a72_test_dl_algo_host_rt.out s:/opt/resnet18/config \
--netBinFile /opt/resnet18/tidl_net.bin \
--ioConfigFile /opt/resnet18/tidl_io_buff1.bin \
--inData /opt/resnet18/in_data_list.txt --inFileFormat 2 \
--outData /opt/resnet18/jet_tidl_out.bin --postProcType 1
```

--------------------------------

### Run TIDLRT CPP Classification Example

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/tidlrt_cpp/README.md

Example command to run the TIDLRT C++ classification example. This command specifies the label file, input image, model artifacts directory, and a debug level.

```bash
./bin/Release/tidlrt_clasification -l test_data/labels.txt -i test_data/airshow.jpg  -f model-artifacts/cl-tfl-mobilenet_v1_1.0_224/artifacts/ -d 1
```

--------------------------------

### Run TIDLRT CPP Examples

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/tidlrt_cpp/advanced_examples/README.md

Executes the compiled TIDLRT C++ application. The example demonstrates priority scheduling and allows for disabling result analysis and specifying test duration.

```bash
./bin/Release/tidlrt_priority_scheduling
```

```bash
./bin/Release/tidlrt_priority_scheduling -r 1 -t 1
```

--------------------------------

### Download and Run Pre-compiled Model

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/README.md

This snippet demonstrates the steps to download a pre-compiled model from the model zoo, prepare its directory, untar the archive, and execute the model using the tfl_main executable with specified input and configuration parameters.

```bash
wget http://software-dl.ti.com/jacinto7/esd/modelzoo/08_00_00_05/modelartifacts/8bits/cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite.tar.gz
mkdir cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite
mv cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite.tar.gz cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite
cd cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite
tar -xvf cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite
cd ../
./bin/Release/tfl_main -z "cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite/" -v 1 -i "test_data/airshow.jpg" -l "test_data/labels.txt" -a 1 -d 1
```

--------------------------------

### Download and Run Pre-compiled Model from Model Zoo

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/osrt_cpp/advanced_examples/README.md

Demonstrates how to download a pre-compiled model from the model zoo, extract it, and prepare it for execution. This involves using wget to download the tarball, creating a directory, moving the file, and then untarring its contents.

```bash
wget http://software-dl.ti.com/jacinto7/esd/modelzoo/08_00_00_05/modelartifacts/8bits/cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite.tar.gz
mkdir cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite
mv cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite.tar.gz cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite
cd cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite
tar -xvf cl-0000_tflitert_mlperf_mobilenet_v1_1.0_224_tflite
```

--------------------------------

### CMake Project Setup and Processor Detection

Source: https://github.com/texasinstruments/edgeai-tidl-tools/blob/master/examples/CMakeLists.txt

This snippet shows the initial CMake setup, including minimum version, project naming, and the logic for detecting the host processor (x86 or ARM). It also handles the critical step of checking for and setting the target device (SOC) and target CPU, with error handling for missing environment variables.

```cmake
cmake_minimum_required(VERSION 3.0.2)

project(edgeai_tidl_examples)

exec_program(pkg-config ARGS --cflags glib-2.0 OUTPUT_VARIABLE GLIB_FLAGS)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${GLIB_FLAGS}")

if(NOT DEFINED CMAKE_SYSTEM_PROCESSOR OR CMAKE_SYSTEM_PROCESSOR STREQUAL "")
  message(WARNING "CMAKE_SYSTEM_PROCESSOR is not defined. Perhaps CMake toolchain is broken")
endif()

message(STATUS "Detected processor: ${CMAKE_SYSTEM_PROCESSOR}")
if(CMAKE_SYSTEM_PROCESSOR MATCHES "amd64.*|x86_64.*|AMD64.*")
  set(HOST_CPU x86)
  if(NOT DEFINED ENV{SOC} OR ENV{SOC} STREQUAL "" )
    message(FATAL_ERROR "SOC not specicfied, please export SOC variable")
  else()
    set(TARGET_DEVICE $ENV{SOC})
    message(STATUS "TARGET_DEVICE setting to: ${TARGET_DEVICE}")
  endif()
  if(NOT DEFINED TARGET_DEVICE)
    message(FATAL_ERROR "SOC not specicfied, please export SOC variable")
  endif()
  if(NOT DEFINED TARGET_CPU)
    set(TARGET_CPU x86)
    message(STATUS "TARGET_CPU not specicfied using ${TARGET_CPU} ")
  endif()
elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64.*|AARCH64.*|arm64.*|ARM64.*)")
  set(HOST_CPU arm)
  if(NOT DEFINED ENV{SOC} OR ENV{SOC} STREQUAL "" )
    message(FATAL_ERROR "SOC not specicfied, please export SOC variable")
  else()
    set(TARGET_DEVICE $ENV{SOC})
    message(STATUS "TARGET_DEVICE setting to: ${TARGET_DEVICE}")
  endif()
  if(NOT DEFINED TARGET_DEVICE)
    message(FATAL_ERROR "SOC not specicfied, please export SOC variable")
  endif()
  if(NOT DEFINED TARGET_CPU)
    set(TARGET_CPU arm)
    message(STATUS "TARGET_CPU not specicfied using ${TARGET_CPU} ")
  endif()
endif()
```