Pose estimation localizes anatomical, robotic, or object keypoints and their spatial relationships in images or video. Human pose supports sports, fitness, ergonomics, sign language, animation, AR/VR, safety, medical analysis, and human-robot interaction; object pose supports manipulation and inspection. 2D pose predicts image keypoints, 3D pose estimates coordinates relative to camera or body, and six-degree-of-freedom object pose estimates translation and rotation. Visibility, skeleton topology, coordinate frame, multi-person association, and uncertainty must be specified. A production perception claim specifies the sensor, scene distribution, label ontology, spatial and temporal resolution, operating range, latency deadline, target hardware, confidence policy, and consequence of a miss or false alarm. Dataset accuracy alone is insufficient when lighting, weather, motion, occlusion, calibration, geography, demographics, and sensor aging differ from the benchmark.

Architecture, representation, and operating mechanism. Top-down systems detect each person then run a keypoint model; bottom-up systems detect joints globally then group them; single-stage methods regress people and poses jointly. OpenPose uses part-affinity fields, HRNet preserves high resolution, ViTPose uses Transformer features, and MediaPipe emphasizes efficient tracking. Models output heatmaps, coordinate distributions, offsets, or direct coordinates. Decoding finds peaks and refines locations; association links joints to individuals; temporal filtering or tracking stabilizes sequences; 3D lifting uses camera geometry and learned priors. PCK, OKS-based average precision, MPJPE for 3D, angular error, visibility accuracy, identity switches, jitter, temporal lag, multi-person recall, latency, FPS, power, and behavior under occlusion and truncation matter. Cameras, lidar, radar, IMUs, optics, illumination, clocks, mounts, compute, memory, interconnect, thermal limits, middleware, trackers, maps, planning, UI, and human escalation form one system. A faster neural network may not reduce end-to-end latency if decode, transfer, synchronization, or postprocessing dominates. Evaluation reports task quality, calibration, subgroup and condition slices, robustness, tail latency, throughput, memory, power, model size, preprocessing and postprocessing cost, and uncertainty across runs. Leakage-resistant splits separate locations, subjects, devices, and time where needed; confidence intervals and error taxonomies expose whether a headline score represents deployable behavior.

Implementation, hardware, and failure modes. High-resolution inputs improve small joints but cost memory; heatmap supervision gives spatial structure; coordinate regression is compact; flip and scale augmentation, synthetic bodies, motion data, kinematic constraints, bone-length priors, distillation, and quantization improve deployment. Multi-person top-down compute scales with detected people, while bottom-up costs more independently of count. Edge NPUs accelerate backbones; decode, association, tracking, camera transforms, and rendering can dominate CPU latency. Self-occlusion, crowding, loose clothing, unusual bodies, fast motion, motion blur, camera foreshortening, mirrored limbs, cropped people, keypoint taxonomy mismatch, biased training, and depth ambiguity create anatomically plausible but wrong poses. Engineering must include data movement, finite precision, resource contention, numerical or physical limits, error propagation, and deterministic behavior when assumptions are violated. The pipeline includes sensing, synchronization, calibration, ingestion, annotation, augmentation, training, evaluation, compilation, quantization, serving, monitoring, feedback, rollback, and dataset/model retirement. Raw data, labels, ontology versions, transforms, checkpoints, compiler artifacts, thresholds, and hardware profiles are traceable so a field failure can be reproduced.

Evaluation, verification, and deployment. Evaluate people- and scene-separated clips, crowd density, body diversity, clothing, camera angle, motion, occlusion, low light, temporal jitter, downstream action impact, and target-device full-pipeline latency. For 3D, coordinate alignment conventions must be reported. Pose feeds activity recognition, biomechanical metrics, avatars, gesture control, safety zones, and robot planning. Camera calibration, synchronization, floor plane, smoothing, latency, and privacy affect downstream meaning. Body and activity signals can reveal health, disability, identity, and behavior. Consent, minimization, on-device processing, retention, access, demographic performance, and non-surveillance alternatives matter. Verification combines held-out and out-of-distribution sets, synthetic stress with real validation, adversarial and corruption tests, calibration analysis, edge-case replay, hardware-in-the-loop timing, long-duration soak, human review, and shadow or canary deployment. Failures feed collection and labeling rather than being hidden by aggregate averages. The pipeline includes sensing, synchronization, calibration, ingestion, annotation, augmentation, training, evaluation, compilation, quantization, serving, monitoring, feedback, rollback, and dataset/model retirement. Raw data, labels, ontology versions, transforms, checkpoints, compiler artifacts, thresholds, and hardware profiles are traceable so a field failure can be reproduced. Evaluation reports task quality, calibration, subgroup and condition slices, robustness, tail latency, throughput, memory, power, model size, preprocessing and postprocessing cost, and uncertainty across runs. Leakage-resistant splits separate locations, subjects, devices, and time where needed; confidence intervals and error taxonomies expose whether a headline score represents deployable behavior.

<svg viewBox="0 0 760 470" xmlns="http://www.w3.org/2000/svg" font-family="-apple-system,Segoe UI,Roboto,sans-serif"><rect width="760" height="470" fill="#0d1117"/><defs><marker id="arrow" viewBox="0 0 10 10" refX="8" refY="5" markerWidth="6" markerHeight="6" orient="auto-start-reverse"><path d="M0 0L10 5L0 10Z" fill="#60a5fa"/></marker><filter id="glow"><feGaussianBlur stdDeviation="7"/></filter></defs><text x="380" y="34" fill="#e6edf3" font-size="21" font-weight="700" text-anchor="middle">Pose Estimation — Heatmaps to a Human Skeleton</text><text x="380" y="56" fill="#8b98a5" font-size="13" text-anchor="middle">joint likelihood peaks connect into an anatomically consistent pose</text><rect x="160" y="78" width="440" height="346" rx="8" fill="#101722" stroke="#3a4453"/><ellipse cx="380" cy="167" rx="58" ry="52" fill="#60a5fa" opacity=".07" filter="url(#glow)"/><ellipse cx="312" cy="220" rx="48" ry="44" fill="#34d399" opacity=".08" filter="url(#glow)"/><ellipse cx="448" cy="220" rx="48" ry="44" fill="#34d399" opacity=".08" filter="url(#glow)"/><ellipse cx="286" cy="307" rx="42" ry="40" fill="#f59e0b" opacity=".08" filter="url(#glow)"/><ellipse cx="474" cy="307" rx="42" ry="40" fill="#f59e0b" opacity=".08" filter="url(#glow)"/><g stroke="#93c5fd" stroke-width="7" stroke-linecap="round"><path d="M380 168L380 258L330 321L310 399M380 258L430 321L454 399M380 205L312 220L278 305M380 205L448 220L480 305"/></g><g fill="#0d1117" stroke="#38bdf8" stroke-width="4"><circle cx="380" cy="140" r="22"/><circle cx="380" cy="205" r="8"/><circle cx="380" cy="258" r="8"/><circle cx="312" cy="220" r="8"/><circle cx="448" cy="220" r="8"/><circle cx="278" cy="305" r="8"/><circle cx="480" cy="305" r="8"/><circle cx="330" cy="321" r="8"/><circle cx="430" cy="321" r="8"/><circle cx="310" cy="399" r="8"/><circle cx="454" cy="399" r="8"/></g><circle cx="480" cy="305" r="16" fill="none" stroke="#fbbf24" stroke-width="1.5" stroke-dasharray="4 3"/><text x="514" y="310" fill="#fbbf24" font-size="10.5">0.61</text><text x="380" y="104" fill="#8b98a5" font-size="10.5" text-anchor="middle">nose · shoulders · elbows · hips · knees · ankles</text><g transform="translate(615 128)"><circle cx="10" cy="10" r="8" fill="#38bdf8" opacity=".3"/><circle cx="10" cy="10" r="3" fill="#38bdf8"/><text x="26" y="14" fill="#8b98a5" font-size="10">heatmap peak</text><path d="M10 43h34" stroke="#93c5fd" stroke-width="5"/><text x="52" y="47" fill="#8b98a5" font-size="10">limb</text></g><path d="M160 352C116 352 116 278 160 278" fill="none" stroke="#a78bfa" stroke-width="1.5" marker-end="url(#arrow)"/><text x="92" y="310" fill="#c4b5fd" font-size="10" text-anchor="middle">spatial</text><text x="92" y="324" fill="#c4b5fd" font-size="10" text-anchor="middle">consistency</text><text x="380" y="452" fill="#6b7684" font-size="11.5" text-anchor="middle">Heatmaps localize joints; skeleton constraints resolve which peaks belong together and which poses are plausible.</text></svg>

Selection and practical application. Top-down models suit accuracy with few people, bottom-up models suit crowds, regression suits compact latency, and temporal/3D systems suit motion analysis when cameras and calibration support them. Coaching, rehabilitation, gesture interfaces, motion capture, workplace ergonomics, collaborative robots, sports broadcast, and sign-language research use keypoint skeletons. Cameras, lidar, radar, IMUs, optics, illumination, clocks, mounts, compute, memory, interconnect, thermal limits, middleware, trackers, maps, planning, UI, and human escalation form one system. A faster neural network may not reduce end-to-end latency if decode, transfer, synchronization, or postprocessing dominates. A production perception claim specifies the sensor, scene distribution, label ontology, spatial and temporal resolution, operating range, latency deadline, target hardware, confidence policy, and consequence of a miss or false alarm. Dataset accuracy alone is insufficient when lighting, weather, motion, occlusion, calibration, geography, demographics, and sensor aging differ from the benchmark. CFS connects this topic to semiconductor architecture, implementation, verification, manufacturing, packaging, test, and deployed AI-system tradeoffs across the platform.