This guide describes how to open and use the Vision AI Setup Assistant information overlay tool. This tool is designed to assess your camera setup for compatibility with the Vision AI feature by providing a set of graphical scores indicating the quality of your video configuration (as indicated by red text for poor, orange text for marginal, and green text for good).
Use these scores to ensure the best possible Vision AI detection during play. This tool assists with positioning cameras at both ends of the ground.
The assistant tool is accessed from the Video Display panel by clicking the Settings cog (top-right) and then clicking Show Vision AI Setup Assistant from the menu.
Once activated, the Vision AI feature begins detecting and tracking objects. If pitch corner detections are successful, a pitch map overlay is displayed. The Vision AI Setup Assistant scoring overlay then appears at the top-left of the Video Display panel, and displays the following information:
Camera Configuration Scores
Detection Scores
Camera Configuration Scores
The camera configuration scores shown in the Vision AI Setup Assistant overlay are:
Video Resolution
The Video Resolution score measures the suitability of the video resolution in use for Vision AI, which works best with high-resolution video footage, preferably 1080p (1920x1080). We do not recommend using a resolution lower than 1080p. If the Video Resolution score is low (indicated by orange or red colouring), increasing the video resolution and/or bitrate is strongly recommended to improve results.
Camera Stability
The Camera Stability score measures the amount of movement in the video frame. A stable camera is essential for achieving good ball tracking and trajectory plotting results, so any camera should be securely anchored to prevent motion, such as swaying in the wind. If the camera motion is minimal, the score will be high, allowing Vision AI to produce accurate tracking results. If unstable (e.g., swaying in the wind), the score will drop, necessitating additional stabilisation.
Note that if monitored during play, the score may drop when players move within the frame. However, assessing Camera Stability when the batter is taking strike provides a more accurate indication of video stability.
Zoom
The Zoom score measures the suitability of the camera’s zoom or distance from the action. Vision AI assesses key object detections and structure sizes relative to known reference sizes. If objects meet or exceed the reference values, the score remains high, ensuring more accurate and reliable results.
If the Zoom score is consistently low (orange or red), increasing the zoom level or moving the camera closer is strongly recommended.
Pitch Field of View
The Pitch Field of View score evaluates the camera's vertical field of view. For optimal detection, the camera should be positioned high enough to capture a large portion of the pitch with minimal obstructions (e.g., umpires or bowlers). A good Pitch Field of View enables accurate ball tracking and trajectory plotting.
For the best results, the pitch should occupy up to 50% of the camera’s vertical field of view. If physically raising the camera is not feasible, consider slight lateral shifts to improve visibility.
The following images show two camera setups, with the red arrows indicating the extent of the pitch, resulting in one with a good Pitch Field of View score (left) and the other with an average score (right). The pitch on the left clearly has more visible area while the pitch on the right is obscured by the umpire, significantly affecting ball and striker detections:
When it is not possible to raise the camera height to see over the umpire, fewer balls will be processed.
Pitch Geometry
The Pitch Geometry score assesses how well the pitch corners are detected and whether the resulting pitch area aligns with expected dimensions. Vision AI performs best when viewing the pitch from behind the bowler and umpire, where the pitch appears trapezoidal. If pitch corners are obstructed (e.g., by players or umpires), the score will drop.
If pitch corners are visible but the Pitch Geometry score is low, the following changes may improve detections:
- Increase camera zoom
- Move the camera closer
- Adjust the camera’s angle relative to the pitch
- Increase video resolution and contrast
Reapplying crease line markings can also enhance corner clarity.
A view from behind the umpire is best, but Vision AI has some tolerance for other pitch geometries. The image below shows a camera position with a low Pitch Field Of View score but a still-viable Pitch Geometry score. If the camera cannot be raised to see over the umpire and bowler, it may still be possible to shift the camera slightly to one side to achieve reliable detections.
Detection Scores
The ball detection scores shown in the Vision AI Setup Assistant overlay are:
Bowler Approach Consistency
The Bowler Approach Consistency score evaluates Vision AI's ability to detect and track bowlers during their approach. This can be assessed pre-match by simulating a bowler's approach or by using pre-existing footage. Vision AI checks for sequential detections and logical positioning.
If the score is low, potential issues may stem from camera positioning, zoom levels, or an unusual bowling approach. This score is only displayed when bowler detections occur; otherwise, it remains greyed out as "n/a."
Ball Consistency
The Ball Consistency score assesses how well Vision AI detects and tracks the ball. Like the Bowler Approach Consistency score, this can be tested pre-match by bowling on an adjacent pitch or by reviewing pre-existing footage. Low scores suggest Vision AI is struggling due to poor video quality (low resolution, contrast, or sharpness) or insufficient zoom.
To improve ball tracking, ensure:
- High-resolution video quality
- Adequate contrast
- Good camera zoom
The Ball Consistency score displays "n/a" when no ball is present.
Pitch Corner Consistency Scores
The Top Left, Top Right, Bottom Right, and Bottom Left Consistency scores indicate how well pitch corners are detected. A stable setup with a good field of view yields high, stable scores. If these scores frequently drop without obstructions, Vision AI may struggle with pitch corner consistency. In such cases, improving camera zoom, positioning, angle, or resolution may help. Also consider re-applying crease line markings.
Hardware and Video Considerations
The Vision AI Setup Assistant tool is designed to provide meaningful coding results across various setup scenarios on common laptop and desktop systems. Since the tool heavily relies on deep neural network algorithms, real-time or near real-time use requires a device with sufficient processing power. For detailed specifications, refer to: Vision AI Recommended Minimum Hardware Specifications.
For post-match processing, refer to: Vision AI Audit.
For optimal results, ensure your video footage provides:
- High video quality (good contrast, minimal blur, even saturation, low noise)
- High-resolution (1080p recommended)
- Progressive scan video (avoid interlaced formats like 1080i)
Vision AI is designed to handle some degradation in video quality but will yield the most accurate results with optimal footage quality. Ensuring the best video setup will maximise Vision AI's effectiveness for detecting and tracking ball movement, player positions, and pitch geometry.