To ensure the Sports2D AI pose estimation engine can accurately track your joints and detect your ground contacts, your video must meet specific conditions. Please follow these core requirements.
🎥 1. Core Video Requirements
1 Side view in the sagittal plane
Place the camera exactly perpendicular to the running lane, so you are filmed purely from the side as you run across the frame. Not diagonally, and not from the front or back. Running left-to-right or right-to-left both work.
Note: Rotating the shoulders strongly toward the camera, or arms crossing in front of the body, can hide joints from the tracker.
2 At least 3-4 consecutive steps in view
The analysis measures ground contact time, step frequency and foot placement from your touchdowns, so it needs several full steps in one continuous shot. Frame a section of the track roughly 10-15 metres wide and run straight through it at full speed. The analysis is tuned for the maximum-velocity phase (typically 20-50 m into a sprint, with an upright posture). Footage of the first steps out of the blocks will score differently because the body is meant to lean much more there.
3 Single person & calm background
Ideally, only the sprinter should be in the frame. Avoid other athletes, spectators, officials, hurdles with human-like silhouettes, window reflections or posters featuring people. Pacers or training partners running next to you will confuse the tracking.
4 Fully fixed camera
Do not pan, zoom, tilt or follow the runner. A tripod is strongly recommended; handheld recording is discouraged. The camera must stay completely stationary, because the software detects ground contacts by looking for a foot that is planted and not moving in the image. Avoid digitally stabilized or cropped footage.
5 Full body visible, level ground
Your head, shoulders, arms, torso, pelvis, both legs and both feet must remain in the frame for the whole recorded section. Film on a flat, straight, level surface (track or flat grass) and keep the camera level with the horizon. Avoid curves, slopes and treadmills.
6 Joints visible & unobstructed
Wear fitted sports clothing in a colour that contrasts with the background; baggy shorts or long loose tights hide the knee and hip. Make sure shoes and ankles are not hidden behind tall grass, lane barriers or equipment. Harsh shadows across the legs reduce tracking quality.
⚙️ 2. Technical Camera Specifications
The accuracy of the report depends heavily on the quality of the video file. A sprint ground contact lasts only about 0.08-0.12 seconds, so frame rate matters most. For the most reliable results, match the following recommendations:
| Property | Our Recommendation | Why? |
|---|
| Framerate |
120 fps or higher (240 fps ideal) |
At 30 fps a ground contact is only 3 frames long, so contact times and step timings become very inaccurate. Use your phone's slow-motion mode. |
| Resolution |
Ideally 1080p or higher |
Sharp boundaries make joints and body contours much easier for the AI to detect. |
| Motion Blur |
As little as possible |
Use bright light and a fast shutter speed. Blurry feet and lower legs severely complicate tracking at top speed. |
| Lighting |
Good and even |
Avoid shooting directly into the sun (silhouettes) and deep shadows across the legs. |
| Camera Distance |
Far enough for the full body, close enough for detail |
A sprinter should fill roughly half to two-thirds of the frame height. If the runner appears too small, joint accuracy drops drastically. |
| Camera Height |
Roughly hip height |
Provides a good baseline to limit vertical perspective distortion. |
| Perspective |
Standard lens (1x) |
Avoid strong wide-angle lenses (like 0.5x on phones) to prevent body-proportion distortion. |
| Framing |
Runner stays fully inside the frame |
If an arm or leg leaves the screen, those joint positions cannot be calculated for those frames. |
⚠️ Important Note on Framerate: Sports2D documentation specifically warns that accuracy significantly decreases when analyzing videos recorded at less than 60 Hz. Standard 25 or 30 fps recordings are not suitable for sprint analysis, and the report will show a warning if your video is below 60 fps.
💡 About the benchmarks: The target ranges in your report are based on maximal-velocity sprinting of well-trained athletes. Recreational runners will often fall in the lower tiers, especially for ground contact time and step frequency. Use the tiers to track your own progress over time rather than as a pass/fail grade.