Two-stage detection

Feature Extractor Backbone

Base CNN network (like ResNet, VGG, or EfficientNet) that extracts visual features from the input image, shared between proposal and classification stages.

← Terug