# Comparisons ## Do I need an NVIDIA Jetson if I use an OAK camera? No. OAK cameras run neural networks on the device, so a Jetson is optional. RVC2 OAK cameras work with any host, including a Raspberry Pi or a laptop. OAK4 cameras run applications standalone, with 48 INT8 TOPS for AI on board. Teams pair OAK cameras with a Jetson when the host needs to run additional workloads beyond vision. Learn more: [Deploy to Jetson](https://docs.luxonis.com/hardware/platform/deploy/to-jetson.md) · [OAK Apps](https://docs.luxonis.com/software-v3/oak-apps.md) ## Stereo depth vs time-of-flight (ToF): which should I use? Stereo depth measures disparity between two cameras and works indoors and outdoors, including in sunlight, at ranges of up to about 12 m. Time-of-flight measures how long emitted IR light takes to bounce back, which gives dense depth on textureless surfaces but usually over shorter ranges, and sunlight can wash it out. Luxonis offers both: stereo OAK-D models, and ToF depth on the OAK-D ToF and OAK-FFC ToF 33D for short-range work. Learn more: [ToF depth](https://docs.luxonis.com/hardware/platform/depth/ToF-depth.md) · [Configuring stereo depth](https://docs.luxonis.com/hardware/platform/depth/configuring-stereo-depth.md) ## Stereo camera vs LiDAR: what is the difference? LiDAR measures distance with laser pulses and gives very accurate, sparse points at long range, but it costs more and produces no color image. A stereo camera produces a dense depth map aligned to a color image at lower cost, with accuracy that drops as distance grows. OAK stereo cameras also run AI on the same device, so detection and depth come from one sensor. Learn more: [OAK depth features](https://docs.luxonis.com/overview/toplevel-features/depth.md) ## Stereo depth vs structured light? Structured light projects a known pattern and measures how it deforms, which gives high accuracy at close range but struggles in sunlight and over distance. Active stereo, used in OAK-D Pro cameras, also projects IR dots, but the dots only add texture, and depth still comes from two cameras. This lets active stereo fall back to passive stereo outdoors, where sunlight washes out the projector. Learn more: [IR dot projector](https://docs.luxonis.com/hardware/platform/features/ir-perception/dot-projector.md) ## Stereo depth vs monocular depth estimation? Monocular depth estimation uses a neural network to infer depth from a single image. It produces relative depth that looks plausible but usually lacks true metric scale. Stereo depth measures metric distance directly from the geometry of two cameras. For robotics, measurement, and spatial detection, OAK stereo cameras give real distances in meters. Learn more: [OAK depth features](https://docs.luxonis.com/overview/toplevel-features/depth.md) ## Classical stereo vs neural stereo? Classical stereo matching (the OAK StereoDepth node) runs in dedicated hardware with low latency and power use, and works well on textured scenes. Neural stereo (the NeuralDepth node, running LENS on OAK4) uses a neural network to fill in low-texture areas, repetitive patterns, and difficult lighting, at a higher compute cost. OAK4 also supports neural-assisted stereo, which combines both approaches. Learn more: [StereoDepth node](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/stereo_depth.md) · [NeuralDepth node](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/neural_depth.md) · [Neural-assisted stereo](https://docs.luxonis.com/software-v3/depthai/depthai-components/nodes/neural_assisted_stereo.md) ## OAK-D vs OAK-D Pro: what is the difference? The OAK-D uses passive stereo and needs scene texture and light. The OAK-D Pro adds an IR laser dot projector for active stereo, which improves depth on blank, low-texture surfaces, and an IR flood LED for operation in darkness. Choose the OAK-D Pro for indoor robots, low light, or textureless scenes. Learn more: [OAK-D](hw-catalog://OAK-D) · [OAK-D Pro](hw-catalog://OAK-D Pro) ## OAK-D SR vs OAK-D vs OAK-D LR: which baseline do I need? Pick the baseline by working distance. The OAK-D SR (2 cm baseline) is for close range, with an ideal range of 10 cm to 1 m and below 2% error out to 1.8 m, for tasks such as bin picking and tabletop robotics. The OAK-D (7.5 cm) covers general use from about 70 cm to several meters. The OAK-D LR (15 cm) targets long range, staying below 1.5% error out to 7.5 m, for outdoor and mobile robots. Learn more: [OAK-D SR](hw-catalog://OAK-D SR) · [OAK-D](hw-catalog://OAK-D) · [OAK-D LR](hw-catalog://OAK-D LR) · [Depth accuracy](https://docs.luxonis.com/hardware/platform/depth/depth-accuracy.md) ## RVC2 vs RVC4: which Luxonis platform should I choose? RVC2 (OAK cameras) suits cost-sensitive, low-power deployments where a host is available, running models and stereo depth on a Myriad X. RVC4 (OAK4 cameras) suits standalone deployments and heavier models, with 48 INT8 TOPS for AI, a Qualcomm CPU running Luxonis OS, NeuralDepth, and OAK Apps. DepthAI v3 runs the same pipeline code on both platforms. Learn more: [RVC2 platform](https://docs.luxonis.com/hardware/platform/rvc/rvc2.md) · [RVC4 platform](https://docs.luxonis.com/hardware/platform/rvc/rvc4.md) ## Why run AI on the camera instead of in the cloud? Running AI on an OAK camera removes the round trip to a cloud server, which cuts latency, bandwidth, and per-inference cost. Video never leaves the device unless the application sends it, which helps with privacy requirements. On-camera inference also keeps working when the network drops, which matters in warehouses, farms, and vehicles. Learn more: [Spatial perception](https://docs.luxonis.com/overview/toplevel-features/spatial_perception.md) · [OAK Apps](https://docs.luxonis.com/software-v3/oak-apps.md)