Cross-scene Transfer of Road Disease Classification Model Based on Dynamic Knowledge Distillation Network

There will be various apparent diseases on the road, which will affect the driving safety and cause economic losses, so it is very important to maintain the road. Researchers need to transfer the classification model of road diseases across scenes, but the data of road diseases in the target scene are few and only partially marked, so the existing transfer learning method has poor effect. Aiming at the problem of cross-scene transfer of road disease classification model, the DNN-Road algorithm is proposed in this paper. The experimental results show that the DNN-Road algorithm achieves the best results among the mainstream transfer learning algorithms.