FilterLoss: A Transfer Learning Approach for Communication Scene Recognition
Jiasong Han, Yufei Feng, Xiaofeng Zhong
TL;DR
The paper tackles data scarcity and label imbalance in communication scene recognition by leveraging transfer learning from SHL-2018 to Beijing-2024. It introduces FilterLoss, a Loss Function with Weight Filter that partitions samples into k weight classes and assigns per-class weights, combined with a weight-assignment algorithm, to bias training toward informative samples. When applied to fine-tuning a ResNet-based backbone, FilterLoss outperforms traditional oversampling and undersampling, achieving up to 92.34% accuracy and improved stability with limited data and computation. The approach enables robust, device-agnostic deployment on mobile sensors with minimal data collection, offering practical benefits for personalized, transferable recognition of communication scenes.
Abstract
Communication scene recognition has been widely applied in practice, but using deep learning to address this problem faces challenges such as insufficient data and imbalanced data distribution. To address this, we designed a weighted loss function structure, named FilterLoss, which assigns different loss function weights to different sample points. This allows the deep learning model to focus primarily on high-value samples while appropriately accounting for noisy, boundary-level data points. Additionally, we developed a matching weight filtering algorithm that evaluates the quality of sample points in the input dataset and assigns different weight values to samples based on their quality. By applying this method, when using transfer learning on a highly imbalanced new dataset, the accuracy of the transferred model was restored to 92.34% of the original model's performance. Our experiments also revealed that using this loss function structure allowed the model to maintain good stability despite insufficient and imbalanced data.