Niladri Shekhar Dutt

Results 2 issues of Niladri Shekhar Dutt

While training the network (l2face_trainer), are we fine tuning the [audio net](https://github.com/zhangzjn/APB2FaceV2/blob/c76e1f560f070382004556a943d16299097c2fcb/trainer/l2face_trainer.py#L27)? Or is the pretrained weights good enough to be used on any video?

**Is your feature request related to a problem? Please describe.** EfficientNet is widely used owing to its speed and accuracy. Since EfficientNet uses Swish unlike ResNet which uses ReLU it...