Is it possible to implement popular DCNN archs like AlexNet, GoogleNet, ResNet?

henrygouk / dopt

A numerical optimisation and deep learning framework for D.

https://henrygouk.github.io/dopt/

BSD 3-Clause "New" or "Revised" License

29 stars 2 forks source link

Open aferust opened 4 years ago

aferust commented 4 years ago

if so, choose one of them and please add an image classification example, for instance, covering

loading the trainig images with opencv (I made a binding to opencv).
preprocessing & normalization: subtract mean image etc.
building the particular model (e.g Alexnet, 'cause it is one of the simplest I think) with pooling layers etc.
training the network.
testing it with a given image. I know it is too much work though, and no commits for a while here.

henrygouk commented 4 years ago

Take a look at the examples directory. The CIFAR-100 example uses a wide residual network.

Yes, the project is currently inactive. I'd still like to work on it again at some point, though I don't have any time in the immediate future.