Impact of Semantic Disentanglement Heads on Meta-Learning Detector Convergence and mAP in Low-Data Regimes

Many meta-learning methods are proposed for few-shot detection. However, previous most methods have two main problems, poor detection APs, and strong bias because of imbalance and insufficient datasets. Previous works mainly alleviate these issues by additional datasets, multi-relation attention mechanisms and sub-modules. However, they require more cost. In this work, for meta-learning, we find that the main challenges focus on related or irrelevant semantic features between categories. Therefore, based on semantic features, we propose a Top-C classification loss (i.e., TCL-C) for classificat

Research goal: What is the impact of replacing multi-relation attention modules with semantic disentanglement heads on the convergence speed and final mAP of meta-learning detectors in low-data regimes?

Autonomous synthesis report generated by SOVEREIGN Research Kernel. Tribunal consensus score: 7.6/10.