Step 1: Enter the DLMM environment. Step 2: Enter the corresponding code directory. Step 3: Run python execute.py, the program will be executed automatically one by one, and the execution and results will be printed.
Code description: ①pretrain.py is the pre-training code. The two modes of image and sound are pre-trained separately. ②estimate.py is that after the pre-training, the two modalities measure the reliability and prediction probability of each sample in the target domain, and by the way, evaluate the performance of direct migration after the pre-training. ③Pseudo_label_generation.py is based on the reliability of the two modalities to generate pseudo labels for subsequent incremental learning. ④retrain.py is the pseudo-labels generated by the two modalities, adding pseudo-labels from easy to difficult to retrain and learn. ⑤ test.py is the accuracy of each mode in the target domain again after the incremental learning of the two modes is completed. ⑥acc_fusion.py is the accuracy of fusing two modes according to the reliability. Cogbeacon dataset:https://pan.baidu.com/s/1rqvRFUQGpK3xuZsnuobqsw(code:E9i3)