In computer vision, scale imbalance refers to a challenge where objects of interest in images have dramatically different sizes (scales), and smaller objects are underrepresented or harder to detect compared to larger ones.For example, in an aerial image, a car (small) and a building (large) differ greatly in scale; a model might bias toward detecting large objects while missing small ones. This imbalance can hurt model performance, as standard training may not sufficiently learn features for tiny objectsAddressing scale imbalance can involve techniques like multi-scale training and inference (e.g., using image pyramids or Feature Pyramid Networks), data augmentation by rescaling objects, or specialized network architectures that preserve detail for small objects. Ensuring the model sees objects at various scales during training helps it perform well across the spectrum of object sizes.
Data Selection & Data Viewer
Get data insights and find the perfect selection strategy
Learn MoreSelf-Supervised Pretraining
Leverage self-supervised learning to pretrain models
Learn MoreSmart Data Capturing on Device
Find only the most valuable data directly on devide
Learn More