search

liguge / 1D-Grad-CAM-for-interpretable-intelligent-fault-diagnosis

智能故障诊断中一维类梯度激活映射可视化展示 1D-Grad-CAM for interpretable intelligent fault diagnosis
91 stars 15 forks source link

计算出的cam的output全为0 #5

Open FuSiry opened 1 year ago

FuSiry commented 1 year ago

(741,) [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.

                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                              1. 0.
                                        1. 0.] 使用V1和V2的pytorch代码计算出的CAM输出权重全为0，请问这是为什么？
liguge commented 1 year ago

这个是你的网络有关系吧 只是适用于卷积神经网络。