At home, it is easy to focus on repetition counts and miss how a movement changes as fatigue builds. Range of motion, pace, and joint position may shift even when the exercise still looks familiar.
This article explains a few useful form cues, the general idea behind camera-based pose estimation, and the limits that matter. Pose estimation can support observation, but it cannot decide whether an exercise is medically safe or appropriate for an individual.
Movement Quality Is One Part of the Picture
Comfort and training response can be affected by several factors:
- Load and recovery — how much you do and how quickly you progress
- Movement control — pace, range of motion, and how consistently you can hold a position
- Individual context — prior injury, pain, mobility, and experience
A camera can only estimate visible positions. It cannot assess pain, internal tissue load, medical history, or every factor that makes a movement suitable for someone. Stop if you feel pain and seek qualified advice when needed.
What Is Pose Detection?
Pose estimation is a computer vision technology that analyzes human skeletal keypoints in video or images to determine joint positions and body angles.
In general, a pose-estimation system may:
- estimate visible keypoints such as shoulders, elbows, hips, knees, and ankles
- derive simple positions or angles from those estimates
- apply movement-specific rules for a supported exercise
- produce a count or a limited cue when the camera view is usable
These are estimates, not a complete biomechanical assessment. Lighting, clothing, camera angle, occlusion, device performance, and body position can all change the result.
Common Cues You Can Review
1. Push-ups: Sagging Hips
Error: Hips too high, excessive lumbar arch
What to notice: Whether your trunk position changes as the set becomes harder
Correct: Keep body in a straight line from head to heels, core engaged
2. Squats: Knee Valgus (Knees Caving In)
Error: Knees collapsing inward (X-shape)
What to notice: Whether the knees continue to track in a comfortable direction relative to the feet
Correct: Knees track over toes, hip-knee coordination
3. Planks: Sagging or Lifting Hips
Error: Lower back bending (sagging) or hips too high
What to notice: Whether you can maintain a position and breathe without pain
Correct: Body in straight line, navel drawn toward spine
4. Running: Overstriding
Error: Foot landing ahead of body center of gravity (“braking effect”)
What to notice: Whether the stride feels controlled and repeatable rather than forced
Correct: Short strides, high cadence, land under center of gravity
The Value of AI Pose Detection
Traditional “follow along with video” training has a fundamental flaw: you can’t see yourself.
Mirrors help, but:
- Mirror angles differ from actual viewing angles
- Hard to watch mirror AND feel body simultaneously during movement
- Many form errors aren’t visible in mirrors
Camera-based pose estimation can complement self-review when the movement and camera setup are supported:
- Simple cues — surface selected movement signals during a supported workout
- Consistent reference — apply the same counting rules from one set to the next
- Hands-free counting — reduce the need to tap the screen during a set
- Visible limits — accuracy still depends on lighting, angle, occlusion, and device conditions
What SuperStrive Supports Today
SuperStrive currently offers selectable camera workouts for squats and jumping jacks only. These supported workouts use the Apple Vision framework on device to detect movement and count repetitions.
Push-ups, planks, and standing side leg raises can still be included in plans and workout logs, but the app does not currently provide live camera correction or camera-based counting for them.
Camera guidance is a workout aid, not a medical assessment or a guarantee of perfect form. Use a suitable camera position and follow the on-screen setup for supported movements; seek qualified advice if you have pain or are unsure about technique.
The “Do More” Misconception
Many people think “exercise = doing movements.”
But movement quality is king.
A smaller set performed with control can be more useful than rushing through a larger number. Treat any camera or recording review as one source of feedback, not proof that a movement is safe or suitable for you.
The Bottom Line
Technique, load, fatigue, and individual history can all affect injury risk. No single camera cue can evaluate all of those factors.
Pose detection can add useful cues for supported movements, but it has limits:
- It depends on camera position, visibility, and the movement being supported
- It should complement body awareness and qualified coaching, not replace them
- It cannot diagnose injury or guarantee perfect form
Next time you do push-ups, try recording yourself with your phone.
You’ll find that many forms you thought were “correct” still have room for improvement.
This is Article 6 in our “Science of Exercise” series. To learn more about the dangers of sedentary behavior, read Sitting Is More Dangerous Than You Think. If you want to understand how to systematize your exercise routine, Why Exercise Doesn’t Need Willpower—It Needs a System has the details. To learn which is better for you, HIIT or cardio, Article 7: HIIT vs Cardio has a detailed comparison.