Getting LightlyEdge
You can get LightlyEdge from your contact at Lightly. You will recieve two files:
- LightlyEdge C++ SDK: A compressed archive with the SDK. The file name will be of the form
lightly_edge_sdk_cpp_{platform}_{version}.{ext}
.
- ML models: A TAR archive of the form
lightly_model_{version}.tar
.
Out of the box, we provide either a Linux x86 64-bit build of a shared library or a Windows x86 64-bit build of a shared library. Upon request, we can provide a distribution for:
- Operating systems: Linux, Windows, Mac OS
- Library type: Shared or static
Installation
Uncompress the LightlyEdge C++ SDK to a local folder. Note that the model should stay a tar file.
Linux
Use unzip
to uncompress into lightly_edge_sdk_cpp
folder:
unzip lightly_edge_sdk_cpp_{platform}_{version}.zip
Verify that the folder has the following layout. The actual contents might differ depending on release version:
lightly_edge_sdk_cpp
├── dist
│ └── cmake
│ └── LightlyEdgeSDK
│ └── FindLightlyEdgeSDK.cmake
├── examples
│ ├── 02_first_run
│ │ └── ...
│ ├── 03_similarity_search
│ │ └── ...
│ ├── 04_diversity_selection
│ │ └── ...
│ ├── 05_adaptive_diversity_selection
│ │ └── ...
│ └── 06_object_detection
│ └── ...
├── include
│ ├── lightly_edge_sdk.h
│ └── ...
└── lib
├── liblightly_edge.a
└── ...
Windows
To uncompress, Right click > Extract All.. > Choose a name, e.g. lightly_edge_sdk_cpp
.
Verify that the folder has the following layout. The actual contents might differ depending on release version:
lightly_edge_sdk_cpp
├── bin
│ └── ...
├── dist
│ └── cmake
│ └── LightlyEdgeSDK
│ └── FindLightlyEdgeSDK.cmake
├── examples
│ ├── 02_first_run
│ │ └── ...
│ ├── 03_similarity_search
│ │ └── ...
│ ├── 04_diversity_selection
│ │ └── ...
│ ├── 05_adaptive_diversity_selection
│ │ └── ...
│ └── 06_object_detection
│ └── ...
├── include
│ ├── lightly_edge_sdk.h
│ └── ...
└── lib
├── lightly_edge.lib
└── ...
Optional: GPU Support
GPU support is available for Nvidia GPUs with the following CUDA driver versions:
- ≥525 for Linux
- ≥527 for Windows
To enable GPU support:
- Either install Nvidia CUDA 12 and Nvidia cuDNN 9 on your machine (recommended)
- Alternatively, we can provide all required dynamic libraries in the package
If GPU is not avalable LightlyEdge will default to a CPU.