ZJ

I have solved it. pytorch vision change to 0.3.1 can run

linux下视频输入换成本地视频，结果保存在一个avi中，去掉 /usr/lib/x86_64-linux-gnu/libsqlite3.so编译可以成功，但是一运行就崩

我在[https://github.com/zeusees/HyperFT/issues/10](https://github.com/zeusees/HyperFT/issues/10)提了个问题，后来通过改变金字塔的搜索框大小从7到21，程序就运行起来了。上一帧人脸框均匀采样64个点求光流，请问你实验中这种效果好么？运行没有报错么

@alrightkami hi, Do you have the script export MaskDINO into ONNX,thanks

请问[这里面](https://github.com/tommyMessi/crnn_ctc-centerloss/blob/master/nets/crnn.py#L225)的注释是不是有问题啊，feature的shape和label的shape注释是不是不对(url)

@moto55 have you solved your problem? how to train dataset without landmark

if the training label is organized in the above form,loss will not calculate landmark loss, right? You mean that the effect will become worse after finetune, right?

训练时候也是resize 300x300，按道理这个也形变了，测试时候应该能解决的啊？

when I change the model to mobilenet0.25 and the nodes name of out out1 out2 to output0 586 585.there are many wrong face detected. Have you solved this problem

change [https://github.com/biubug6/Face-Detector-1MB-with-landmark/blob/master/Face_Detector_ncnn/FaceDetector.cpp#L45](https://github.com/biubug6/Face-Detector-1MB-with-landmark/blob/master/Face_Detector_ncnn/FaceDetector.cpp#L45) false, there are also many wrong face detected,but the number is different from above.