Ingredient
NVIDIA Jetson
Also known as: Jetson Nano, Jetson Orin Nano, Jetson Xavier NX, Jetson AGX Orin
NVIDIA's family of GPU-accelerated single-board computers — ARM CPU paired with an integrated CUDA-capable GPU on a credit-card-to-paperback form factor. The right ingredient when computer vision or ML inference is the bottleneck and a Raspberry Pi can't keep up. Models: Jetson Orin Nano ($249, 40 TOPS, 8 GB), Orin NX (100 TOPS), AGX Orin (275 TOPS, $2000). Runs JetPack (Ubuntu + CUDA + cuDNN + TensorRT) out of the box. The default brain for autonomous-rover, weeding-robot, and crop-monitoring projects that need real-time YOLO, semantic segmentation, or SLAM.
Inputs / outputs
- CPU: 6-core ARM Cortex-A78AE (Orin Nano) up to 12-core (AGX Orin)
- GPU: 1024-core Ampere (Orin Nano) up to 2048-core (AGX Orin), with tensor cores
- RAM: 4–64 GB LPDDR5
- I/O: USB-C, USB 3.2, Gigabit Ethernet, M.2 NVMe, 40-pin GPIO header (RPi-compatible), CSI camera (up to 4 simultaneous)
- Power: 7–60 W depending on model and mode (Orin Nano: 7W / 15W modes)
Solves / unlocks
- Real-time crop-and-weed segmentation (YOLOv8 + custom dataset)
- Visual SLAM for in-field navigation
- Pest detection at the leaf level (run inference on every camera frame)
- On-device ML training for small datasets (transfer learning works on Orin Nano)
- Multi-camera vision fusion (4 simultaneous CSI cameras)
Constraints
- Cost — entry-level $249, top-end $2000. Overkill for sensor-logging tasks.
- Power draw — even a 7W mode is too much for solar-only field deployment without a sizable panel.
- Driver weight — JetPack is heavy; image flashing via SDK Manager is finicky.
- Software ARM-only — some ML frameworks are x86-first and need ARM ports.
Source
- Product line: https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/
- JetPack SDK: https://developer.nvidia.com/embedded/jetpack
- Jetson Containers: https://github.com/dusty-nv/jetson-containers (community ML container library)
See also
Auto-generated from this entry’s typed relations: frontmatter, grouped by relation type so the editorial signal isn’t flattened.
- Parallels: [[raspberry-pi]]
- Member of: [[ingredient]]
- Combines with: [[opencv]] · [[ros2]] · [[rgbd-camera]] · [[lidar-rangefinder]] · [[acorn-rover]]
What links here, and how
Inbound connections from across the wiki, grouped by lens and by relationship. These appear automatically — every entity page declares what it links to, and that data populates here on the targets.
Practical
combines with
- Acorn rover Acorn's reference architecture uses Jetson for vision + planning
- GPS-RTK Jetson + RTK-GPS = cm-accurate field navigation
- LiDAR rangefinder LiDAR + Jetson = obstacle-avoidance and crop-row tracking for autonomous farm machines
- OpenCV OpenCV with CUDA backend = real-time YOLOv8/v10 at 30+ FPS on Orin Nano
- PlantVillage dataset Jetson + PlantVillage-trained model = edge-side disease classification at the leaf
- PX4 / ArduPilot autopilot Jetson companion computer attached via UART runs vision and offloads to PX4 for low-level flight control
- RGB-D camera Jetson + RealSense = real-time visual SLAM and 3D perception for autonomous farm machines
- ROS 2 the standard ROS2 perception-and-navigation node for hobbyist autonomous rovers
- YOLO (object detection) YOLOv8/v10 runs at 30+ FPS on Orin Nano with TensorRT optimization
contains
- Farm-tech toolkit compute / GPU-accelerated SBC for vision and ML
parallels
- Raspberry Pi Jetson is the GPU-accelerated step up when ML inference is the bottleneck
combines
- Recipe: autonomous row-crop weeder perception compute — runs YOLOv8 inference at 30 FPS
- Recipe: NDVI crop-scouting drone optional: companion compute on-drone for real-time vision; or post-process on ground
13 inbound links · 7 outbound