can it convert the undefine format model?

ENCP / CNNdroid

Open Source Library for GPU-Accelerated Execution of Trained Deep Convolutional Neural Networks on Android

MIT License

539 stars 181 forks source link

can it convert the undefine format model? #27

Open eeric opened 7 years ago

eeric commented 7 years ago

sush as following: layer{ name: "slice_fc1" type:"Slice" slice_param { slice_dim: 1 } bottom: "fc1" top: "slice_fc1_1" top: "slice_fc1_2" } layer{ name: "etlwise_fc1" type: "Eltwise" bottom: "slice_fc1_1" bottom: "slice_fc1_2" top: "eltwise_fc1" eltwise_param { operation: MAX } } it was from that the light cnn C model: https://github.com/AlfredXiangWu/face_verification_experiment

latifisalar commented 7 years ago

Unfortunately not, implementation of the custom layers needs to be added for the support.

akbarxie commented 7 years ago

Tecent yesterday release their ncnn,contain the layers you need ,but they only runs on the cpu,and not fast enough,but it may meets your needs.

latifisalar commented 7 years ago

That's great, thanks for the update and your help, we'll have a look to their implementation.

eeric commented 7 years ago

@latifisalar, it was to be for the support to slice and etlwise, was it?