Hans Ren

可以参考validation阶段是如何做的。

查看resizeNormalize，如果你输入图片是opencv格式的话，resizeNormalize会损坏图片内容导致输出都为“blank”。 可以查看#213

@edubois For face rotation, I advise you to try PCN. This network is similar to mtcnn.

@bonseyes-admin @wywu could someone close this issue please?

@yyfanxing 请问您解决这个问题了嘛？我最近在用这个模型训练，也发现不收敛。

> I tried authorizing my client using: **_./tailscale -socket var/run/tailscale/tailscaled.sock up_** and then it gives me a message saying **_Failed to connect to local tailscaled; it doesn't appear to be...

@qingzew did u solve this problem?

I solved this problem by adding **engine: CAFFE** under **convolution_param** in every convolution layers in train_val.prototxt

> Hey @DentonGentry , FWIW, after running : > > ``` > sed -i -e 's/QTS_MINI_VERSION=".*"/QTS_MINI_VERSION="4.0.0"/' ./Tailscale/qpkg.cfg.in > ``` > > and > > ``` > make pkg > ```...

> make build-qdk-container @DocEmmetBrown Brilliant! I have had them run properly. Thank you, Emmet!