Efficiency between tensorflow and caffe?

hx173149 / C3D-tensorflow

C3D is a modified version of BVLC tensorflow to support 3D ConvNets.

MIT License

588 stars 262 forks source link

Efficiency between tensorflow and caffe? #17

Closed LongLong-Jing closed 7 years ago

LongLong-Jing commented 7 years ago

@hx173149 Did you notice the difference in efficiency between this code and original C3D code? During testing, the batch size of caffe code can be 30 in GTX, while the batch size for this code is 15. When adding the batch size, the code will report 'Out of Memory' error. Can you tell me you GPU configuration and batch size during training? Thanks!

hx173149 commented 7 years ago

Hi @LongLong-Jing Yes, maybe there have some points to optimize the efficiency.