All skills
deepstream-import-vision-model
Import an object-detection model into a DeepStream pipeline and produce a benchmark PDF report.
- Part of NVIDIA/skills
- 2.3k installs
- Machine Learning · Vision models
When to use it
You have a HuggingFace or NGC detection model and need a working TensorRT engine, nvinfer config, and benchmark numbers.
For example
You say
Use deepstream-import-vision-model to bring PekingU/rtdetr_r50vd into a DeepStream pipeline and benchmark it.
You get
- TensorRT engine file
- nvinfer config and bbox parser
- Benchmark PDF report
How it works
- Only object-detection models are accepted; classification and segmentation are rejected after checking config.json.
- SafeTensors models without a published ONNX are exported first, with a check that the batch dimension stays dynamic.
- A KITTI detection check must reach at least 90% detection rate before multi-stream benchmarking.
Before installing
Needs Docker with the NVIDIA driver and a GPU; everything runs inside the nvcr.io/nvidia/deepstream:9.1-triton-multiarch container.
This skill includes 33 scripts your assistant may run on your computer.
More Vision models skills
See allAlso in this pack
- accelerated-computing-cudf3.2kGuides GPU DataFrame work with cuDF, including pandas acceleration, joins, groupby, file I/O and multi-GPU jobs.
- cuopt-developer2.8kModifies, builds, tests, and debugs NVIDIA cuOpt source code and prepares contributions.
- cuopt-install2.6kInstalls cuOpt for Python, C, or server via pip, conda, or Docker and verifies it works.
- data-designer2.6kGenerates a synthetic dataset from your description of the data you need.