Split - Ratio Balance

In QpiAI Pro, the Split feature is a data partitioning and balancing mechanism that intelligently divides your dataset into Training, Validation, and Test sets based on defined ratios or intelligent auto-balancing.Unlike random splitting, which may cause data imbalance (e.g., some classes or scenes appear more in validation than training), the QpiAI Split ensures your data distribution is optimized for performance and model generalization.

When using the Split feature, you can: Choose the Split Mode:

• Skip → The system uses the default backend split (predefined 70/20/10)
• Rebalance Ratio → You define how you want your data divided (custom ratio). Define Ratios→ Train: 70%, Validation: 20%, and Test: 10% or Your choice.