Vision

Function Summary

Using the Vision function, the position of each joint in the body (skeleton) can be estimated from the video, and changes in angles and the magnitude of movement can be quantified and visualized.
Even if there is no sensor data (even if there is no TYPE-S), there is a possibility to analyze the movement just by taking video. Of course, with sensor data, a deeper analysis will be possible.

To evaluate "thigh lift" on a run or bike, based on the angle of the thigh to the horizontal
Adjust position while pedaling a bicycle, taking care not to over-extend the knee joint angle
Evaluate depth based on thigh angle in squat
Show the trajectory of head movement and check for stability
Combine PelvicAngle and Vision thigh angles to estimate hip flexion angle

Overlay toggle icon

Left icon toggles the angle overlay, Right icon toggles the trajectory.
Trajectory toggle

When the trajectory display is turned on, the center of the joint point turns yellow

By tapping each joint, you can switch the display settings for each joint.
- If you have difficulty tapping, pause the video, zoom in on the image, or move frame by frame.
Angles and trajectory can be displayed for each joint.
- V3.0 can display "Knee joint angle", "Thigh angle", and "Shin angle" as angles
- Trajectories can be displayed for all joints

Angle information can be viewed as a Graph.
Select the "Vision" tab in the Graph Item Settings window and choose the angle information you wish to display.
Only captured data is displayed. in Capture mode, the graph is updated in real time, so there may be some flickering.

Mode
- View mode: Plays back captured (recorded) data. No video processing is executed. (Thus, smoother movement and lower battery consumption)
- Capture mode: Processes video in real time and saves skeletal data sequentially to a file. (Requires a more powerful processor and consumes more battery power)
Threshold
- "Low" is the basic setting. It is not necessary to change it if the behavior is not strange.
- In rare cases, a skeleton that does not exist may be detected because it is misled by the background or a moving object. In such cases, increase the Threshold to tighten the detection threshold. This will reduce the amount of noise data.
"Clear Data" button
- In Capture mode, you may want to change the Threshold and reprocess the data even if it has already been captured. In such a case, delete all captured data from this button and re-capture.

When I try to change to Capture Mode, I get an alert and can't change it.
- You can only use Capture mode on devices with iOS 14/iPadOS 14 or higher and A12 or higher SoC (iPhone XS or later)
I selected a graph but nothing is displayed.
- Simply opening a file does not process the skeletal data and the data will not appear on the graph. Once the file is in Capture mode and played back, the data will be saved and the graph will be drawn.
Zooming in on the graph shows missing data (rough data density)
- Since processing in Capture mode is in real time, data may be missing in slower SoC models. Please decrease the playback speed or move the red line to the missing area and perform Capture processing.
I want to do a training while watching angles in real time.
- Vision function is not available during recording. You can use it on the pre-recorded screen, so you can do a training while watching it.

If there is more than one person on the screen, the person in nearest is processed.
Vision function is not available while recording (skeleton will not be displayed).
Capture Mode is only available on devices with A12 or later chips
In V3.0, angles can only be displayed with limited joints
Video output via Export is not supported.
Capture mode is not available in Compare.