RCCM: Reinforce Cycle Cascade Model for Image Recognition
Hongjun Li, Ze Zhou, Chaobo Li, Shibing Zhang · IEEE Access · 2020
Due to the disadvantages that complex network structure, time-consuming training process and the insufficient feature extraction ability existing in deep structures and broad structures, Reinforce Cycle Cascade Model (RCCM) is proposed to offer a novel algorithm for image recognition. In RCCM, the cycle cascade model is used to extract the discriminative features with a small amount of time consumption, where the multi-layer cascade is used to extract low-level and advanced features layer by layer, which provides conditions for features evolution as well. In addition, the cycle mechanism is introduced to reinforce features gradually, which also lighten model complexity by adopting time update instead of space stack. Visualization maps prove the feasibility of RCCM, and the discussion of cascade layers shows the extension of our proposed algorithm. To demonstrate the performance of RCCM, we have conducted extensive experiments on some benchmark image datasets, and the results show that RCCM is significantly better than several state-of-the-art algorithms. Specially, no matter in the training process of testing phase, our method can achieve recognition in real-time even on common PC without GPU, and the maximum testing speed reaches 31073 samples per second. RCCM improves the efficiency and accuracy of image recognition significantly.