search

tensorflower / seetaFace6Python

简单、快速搞定人脸识别应用，觉得有帮助，给个start吧！
198 stars 54 forks source link

docker部署并发超过2个,貌似内存泄漏,服务会重启, #53

Open yangjian1218 opened 3 years ago

yangjian1218 commented 3 years ago

docker部署,使用了人脸检测跟活体检测,我在使用常规python的活体检测,并发没有问题.但使用seetaface的动态库,并发超过2个,就会导致服务蹦,进程重启,好像是内存泄漏导致的? 请问这个情况怎么解决? `Stack trace: [bt] (0) /usr/local/lib/python3.6/dist-packages/mxnet/libmxnet.so(+0x2f9cf20) [0x7fe087731f20]

[bt] (1) /lib/x86_64-linux-gnu/libc.so.6(+0x3f040) [0x7fe1a3812040] [bt] (2) /home/model-server/live_flask/seetaface/lib/ubuntu/libSeetaFaceLandmarker600.so(+0x10c51) [0x7fdfeaf1fc51] [bt] (3) /home/model-server/live_flask/seetaface/lib/ubuntu/libSeetaFaceLandmarker600.so(seeta::v6::FaceLandmarker::mark(SeetaImageData const&, SeetaRect const&, SeetaPointF*) const+0x29) [0x7fdfeaf20cb9] [bt] (4) /home/model-server/live_flask/seetaface/lib/ubuntu/libFaceAPI.so(mark5+0x3a) [0x7fdff069ecaf] [bt] (5) /usr/lib/x86_64-linux-gnu/libffi.so.6(ffi_call_unix64+0x4c) [0x7fe19e063dae] [bt] (6) /usr/lib/x86_64-linux-gnu/libffi.so.6(ffi_call+0x22f) [0x7fe19e06371f] [bt] (7) /usr/lib/python3.6/lib-dynload/_ctypes.cpython-36m-x86_64-linux-gnu.so(_ctypes_callproc+0x2b4) [0x7fe19e2775a4] [bt] (8) /usr/lib/python3.6/lib-dynload/_ctypes.cpython-36m-x86_64-linux-gnu.so(+0x11d34) [0x7fe19e277d34] [2021-08-26 10:58:35,583] INFO [start_seeta_2d.py:240] model is all loaded [2021-08-26 10:58:35,697] INFO [glogging.py:264] Booting worker with pid: 2098 `

yangjian1218 commented 3 years ago

报错类型为Segmentation fault: 11, 这类报错为指针问题.

w-Bro commented 2 years ago

@yangjian1218 我多线程并发使用就会有这个问题，请问您解决了吗？

yangjian1218 commented 2 years ago

@yangjian1218 我多线程并发使用就会有这个问题，请问您解决了吗？ @w-Bro 后来我自己去写了源码编译,用我自己编译的了,如果你想用,去我主页看.