Anders Xiao

顺便提供一个编译好的 windows onnx runtime 静态库 (请自行下载 onnx runtime 1.15.1 动态库)，方便后来的同学使用~ PS：刚处理好编译问题，还没来得及测试 [funasr_winx64_static_md.zip](https://github.com/alibaba-damo-academy/FunASR/files/12009971/funasr_winx64_static_md.zip)

@csukuangfj K2那个乱码的问题我觉得真的是K2的问题，如果要调试源码那就要自己维护分支了，所以我改用FUNASR了，目前一切正常

@csukuangfj 那回头我再调试一下，另外标点恢复真的很需要，是不是考虑训练一个，转到FunASE一部分原因它自带了标点恢复模型。

@csukuangfj 我其实就是用在游戏引擎里，我觉得 NCNN 之类的手机端效率更高，我知道 onnx 和 ncnn 的区别，我是用你们的 ncnn 版本出现问题就没尝试 onnx 版本了，标点模型似乎 K2 官方没有提供，开发者自己去提取其他方案的当然可以，但是我建议自带一个，这东西 FunASR有， Paddle 也有，用 K2 就要自己找一个，然后要配合 VAD 附加标点，那多多少少开发量又有一些了，FunASR 了的 OnnxRuntime 只要指定模型路径，都不用自己操心了，真的开箱即用，拿到的输出就有标点（标点效果一般，但普通业务足够用），这其实大大的减少了用户的开发量，当然了，只是个人建议~

Cool !

@yichunk Hi, thanks for your reply, sorry for late. Here is my graph: ``` node { calculator: "ImageTransformationCalculator" input_stream: "IMAGE:input_video" output_stream: "IMAGE:transformed_image" input_side_packet: "FLIP_HORIZONTALLY:input_horizontally_flipped" input_side_packet: "FLIP_VERTICALLY:input_vertically_flipped" input_side_packet: "ROTATION_DEGREES:input_rotation" } node...

It has to be said that most of the BlendShapes work well, and the problems mainly appear in the eyes. My program calibrated bs before running mediapipe. 1. When the...

@creativewax Hi，I think you'd better wait google, I have done legacy solution (in Unreal Engine), but its not good enough. I also done new solution in Android, and push to...

@creativewax You can use my graph above, but you need dowload face_blendshapes.tflite file. BS is in ClassificationList. Label is bs name , value is bs weight (0-1) ,I saw google...

BTW, I still think it's a non-production-ready solution and its effect is almost unusable because the eye part is not handled well. For a production application，you can consider NVIDIA AR...