Alexandr

Hi, Seems that exactly the same story with me as with @ernestchancivitas (no, we are not working in the same office J). I wonder if there is a way to...

Hi, It is not a secret that **Caffe** generally not efficient in memory usage. It is a pay for speed. cuDNN could slightly improve the situation. Best

@mrgloom: > is it related to convolution implementation Seems yes.

IMHO, integration with **Caffe** is important for growth since **Nvidia** is taking steps towards this framework, such as **Digits** or **GIE** and **SegNet**'s idea and implementation still original and useful.

Hi, Thanks for the response. If I'm right, in FMs there is no explicit batches as in NNs. In a sample-wise splitting, at least two questions come to mind: *...

Hi, Unfortunately, priorities changed rapidly. I nearly had not a chance to handle this issue.

Hi, @fakjhhoetg `frames_*.npy` have the following essence: it is a `dict` of raw frames where the key is a filename and the value is a `numpy.ndarray` with shape `(3, 224,...

Hi, @fakjhhoetg `X*.npy` stores raw frames and `y*.npy` stores corresponding emotion labels. `X*.npy` is 4D array where the channel index really depends on your **Keras** configuration, e.g. `(N, 3, 224,...