Table 3 FL approaches and applications in indoor image classification.
Refs. / Year | Domain | Datasets Used | Techniques/Models | Results |
|---|---|---|---|---|
43 / 2025 | Medical image classification using blockchain-based FL | Not specified | FedBG (blockchain-based FL, EC-GAN) | 27–38% reduced training time, 0.9–2% better accuracy, privacy preserved |
44 / 2024 | Indoor image-based positioning using FL | Not specified | MobileNet, FedAvg, FedOpt | 94% accuracy, client privacy preserved |
45 / 2024 | Efficiency of FL algorithms on image classification | Fashion MNIST | FedAvg, FedBABU, FedProto, APPLE | APPLE: 99.21%, FedProto: 99.40% accuracy (non-IID data) |
46 / 2021 | Bridging P-FL and G-FL for image classification | Diverse datasets | FED-ROD framework | Superior scalability, balanced accuracy with non-IID data |
47 / 2023 | FL for medical image analysis | Not specified | Deep Neural Networks | Comparable results to centralized models, privacy preserved |
48 / 2022 | FL for breast cancer histopathological image classification | BreakHis | DenseNet-201, ResNet-152 | Comparable to centralized models, high reliability and consistency |
49 / 2022 | FL with blockchain for UAV image classification | Not specified | FLBIC-CUAV (FL, blockchain, ResNet, BSO) | 99.15% accuracy, reduced delay, improved throughput, lower energy consumption |
