computervisionlearner

sphereFace划分的类别数等于训练集中人脸总数吗？

1

之前一直在研究局部描述子特征匹配，发现与人脸识别有很多相似之处，所以想借鉴一下。对于Siamese和Triplet这种度量方式很好理解，但是对于改进softmax这类方法，存在一处疑惑。sphereFace划分的类别数就是数据集中人脸的总数吗？

Hi Dear: I have read the paper that you published in arxiv (https://arxiv.org/abs/2410.03930).  As descripted above, I am confused that what's the difference between `Reverb` and `Reverb Research`. Why...

refactor moe of positionwise_feed_forward to let us convert onnx success

6

对Moe的实现进行了重构，原来的Moe实现带有python的流程控制，会导致导出onnx失败。修改后的moe实现，通过mask可以让pytorch更好动态追踪expert的选择。

## 🐛 Bug ## To Reproduce Steps to reproduce the behavior: 1. set max_speech_duration_s = 20 in https://github.com/snakers4/silero-vad/blob/master/examples/cpp/silero-vad-onnx.cpp. like this: ``` #include #include #include #include #include #include #include #include #include...