ViMo SDK Introduction

The ViMo SDK is part of the SMore ViMo platform. It builds on SmartMore’s platform and AI capabilities to provide tools for local model execution. Users can design solutions in SMore ViMo, export SDKs, and enable third‑party development for detection, classification, segmentation, single‑character OCR, and sequence‑level OCR across diverse industrial vision scenarios. This allows customers to focus on business value while using stable and efficient AI.

What’s new
[3.14]
1. Pipelines support parameter editing
[3.13]
1. Added multi‑rotated‑rectangle ROI module
2. Added Fast Positioning module
3. Support for rotated‑rectangle ROI input
[3.12]
1. Batch inference in Pipelines mode
2. TRT cache path aligned to model path
3. Dynamic masking for segmentation and unsupervised modules
4. Fixed AnyResponse empty ROI issue
5. Coefficient‑region setting in Unsupervised module
[3.11]
1. Visualisation supports Chinese
2. Segmentation masks and region‑level thresholds can be based on viewport or full image
[3.10.0]
1. Class‑agnostic NMS option for detection and rotated‑rectangle detection
2. Batch inference in single‑module mode
3. Set maximum batch size on Module/Pipelines initialisation

Development environment requirements
Development environment requirements
1. Device
- CPU: Intel Core i7 or above
- Memory: 8GB or more
- GPU (for GPU mode): NVIDIA GPU with compute capability 6.0 or higher
2. GPU environment
- CUDA 11.4
- cuDNN 8.2.4
- NVIDIA driver 471.41 or later
Notes:
- If your CUDA version exceeds 11.4, ensure the driver supports it.
- See Appendix: Configure NVIDIA GPU driver and GPU dependencies.
3. TensorRT
- Install TensorRT (recommended 8.6.1 or later) in addition to GPU environment.
- See Appendix: Configure TensorRT dependencies.
4. Windows
- Operating system: Windows 10
- Visual Studio 2017 or newer
Notes: Microsoft Visual C++ Redistributable required
5. Linux
- Operating system: Ubuntu 18.04
- CMake 3.16 or newer