Skip to main content
. 2019 Mar 18;19(6):1356. doi: 10.3390/s19061356

Figure 12.

Figure 12

Performance comparison of different models in accuracy using original 51D features when setting Tx to 60 samples and Δt to 6 samples. (a) Performance comparison in the training stage. (b) Performance comparison in the validation stage. Here, the attention-based DeepConvGRU and DeepConvLSTM also yielded good performance.