#6565th of 65 in Model serving
deepstream-yolo
NVIDIA DeepStream configs and GPU parsers for running YOLO and DETR detectors
- Stars
- 2.1k
- License
- MIT
- Last commit
- Jan 2026
- Language
- Python
1,975 on 21 Apr 20262,079 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Configuration and custom TensorRT parsing library (nvdsinfer_custom_impl_Yolo) for running YOLO-family detectors inside NVIDIA DeepStream, from SDK 5.1 to 8.0 on x86 and Jetson. It converts Darknet cfg/weights or ONNX exports into TensorRT engines and does bbox post-processing on the GPU. It also supports INT8 PTQ calibration, non-square input and dynamic batch size.
Who it is for: Engineers deploying YOLO or DETR object detection on NVIDIA DeepStream pipelines
Strengths
- Covers about 27 detector families, including YOLOv5 to YOLO26, RT-DETR, D-FINE and RF-DETR
- GPU bbox parser, INT8 PTQ calibration, dynamic batch size and non-square models
- Per-model usage docs plus benchmarks and a custom ONNX model guide
- Supports DeepStream 5.1 to 8.0 on both x86 and Jetson
Weaknesses
- Tied to NVIDIA DeepStream, TensorRT and CUDA; no use outside the NVIDIA stack
- Author says updates are currently limited
- TensorRT engine generation can take more than 10 minutes
- Needs manual CUDA_VER matching and a make build of the library; no release published
What it needs
- GPU required
- Needs NVIDIA DeepStream SDK, CUDA, TensorRT, GStreamer, NVIDIA driver, JetPack (Jetson)
- Models: Darknet (YOLOv4, etc.), YOLOv5, YOLOv6, YOLOv7, YOLOv8
Quick answers
Is deepstream-yolo open source?
Yes. Its code is on GitHub under the MIT license.
Does deepstream-yolo need a GPU?
Yes, it needs a GPU.
Also in Model serving
See all 65| Rank | Project | Score |
|---|---|---|
| 1 | llama.cppC/C++ inference engine for GGUF LLMs and VLMs with an OpenAI-compatible server | 94 out of 100 |
| 2 | OllamaLocal model runner with CLI and REST API for open-weight LLMs | 92 out of 100 |
| 3 | LocalAISelf-hosted inference server with OpenAI-compatible API and pluggable backends | 91 out of 100 |
| #4 | vLLMInference engine serving LLMs with PagedAttention and an OpenAI-compatible API | 88 out of 100 |
| #5 | strataRuns a 125B-parameter MoE model on one consumer GPU plus system RAM | 88 out of 100 |