MTT-Road: Dynamic Updating of Road Diseases Classification Model Based on Datasets Distillation
- Resource Type
- Conference
- Authors
- Han, Haihang; Ge, Yongjun; Wang, Xinxiao; Yu, Zhi; Xi, Chenchen; Wang, Yue
- Source
- 2023 IEEE 3rd International Conference on Electronic Technology, Communication and Information (ICETCI) Electronic Technology, Communication and Information (ICETCI), 2023 IEEE 3rd International Conference on. :130-134 May, 2023
- Subject
- Communication, Networking and Broadcast Technologies
Power, Energy and Industry Applications
Robotics and Control Systems
Heuristic algorithms
Roads
Transfer learning
Inspection
Data models
Classification algorithms
Safety
Road diseases
continual learning
Image classification
- Language
The apparent diseases of road will cause great losses to driving safety and economic benefits, so when the apparent diseases of road appear, we need to detect them quickly and accurately. Engineers need to use the classified error data in the inspection process to continually and dynamically update the model to avoid similar errors in the subsequent inspection journey. The existing methods do not balance the resource consumption and model effect in the iterative update of the model. In order to solve the problem of continual dynamic update of road disease classification model, the MTT-Road algorithm is proposed in this paper.The experimental results show that the MTT-Road algorithm achieves the best results among the mainstream transfer learning algorithms considering both the time consumption and model effect.