Research map / Federated learning and privacy / Personalized and heterogeneous FL
Navigating Data Heterogeneity in Federated Learning: A Semi-Supervised Federated Object Detection
Taehyeon Kim, Eric Lin, Junu Lee, Christian Lau, Vaikkunth Mugunthan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use FedSTO instead of FedAvg | YOLOv5 Large SSFOD · BDD100K Non-IID | mAP@0.5: 0.558 -> 0.593 | Improved |
| Use FedSTO instead of FedAvg | YOLOv5 Large SSFOD · BDD100K IID | mAP@0.5: 0.591 -> 0.609 | Improved |
| Add local EMA pseudo-labeling to federated training (ablation) | SSFOD · BDD100K Non-IID | mAP@0.5: 0.511 -> 0.558 | Improved |
| Add selective training after SSFL with Local EMA Model (ablation) | SSFOD · BDD100K Non-IID | mAP@0.5: 0.558 -> 0.567 | Improved |
| Add selective training after SSFL with Local EMA Model (ablation) | SSFOD · BDD100K IID | mAP@0.5: 0.591 -> 0.587 | Hurt |
| Add full-parameter training with orthogonal enhancement (ablation) | SSFOD with selective training · BDD100K Non-IID | mAP@0.5: 0.567 -> 0.593 | Improved |
| Add full-parameter training with orthogonal enhancement (ablation) | SSFOD with selective training · BDD100K IID | mAP@0.5: 0.587 -> 0.609 | Improved |
| Remove selective training (ablation) | FedSTO · BDD100K Non-IID | mAP@0.75: 0.289 vs 0.338 | Hurt |
| Remove selective training (ablation) | FedSTO · BDD100K IID | mAP@0.75: 0.311 vs 0.357 | Hurt |
| Use client-local EMA instead of a global pseudo-label model | Federated SSFOD · BDD100K Non-IID, weather-specific | mAP@0.5: 0.555 -> 0.560; 0.560 -> 0.566; 0.497 -> 0.553; 0.488 -> 0.553 | Improved |
| Use a reduced Phase 1 model that excludes the neck when communicating | FedSTO · 100 clients, 350 rounds | network bandwidth: 2,725.50 GB -> 2,166.23 GB; 20.52% | Improved |
| Increase client sampling ratio from 0.1 to 0.2 | FedSTO · BDD100K, unlabeled Bus | mAP@0.5: 0.573 -> 0.378 | Hurt |
Closest papers by meaning
- (FL)$^2$: Overcoming Few Labels in Federated Semi-Supervised Learning (NeurIPS 2024 · similarity 0.89)
- Exploring One-Shot Semi-supervised Federated Learning with Pre-trained Diffusion Models (AAAI 2024 · similarity 0.89)
- Progressive Distribution Matching for Federated Semi-Supervised Learning (AAAI 2025 · similarity 0.88)
- SemiDFL: A Semi-Supervised Paradigm for Decentralized Federated Learning (AAAI 2025 · similarity 0.87)
- BSemiFL: Semi-supervised Federated Learning via a Bayesian Approach (ICML 2025 · similarity 0.87)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
