Vision: Add Inception V2, Inception V4, Inception Resnet V2 in models

[wanghan0501]

Why not add Inception V2, Inception V4, Inception Resnet V2 in models？？

[VibAltekar]

I'd love to give this a shot! Is anybody already working on this?

[karandwivedi42]

@VibAltekar do go through the discussion here if you want to add a new model: https://github.com/pytorch/vision/issues/321#issuecomment-341820560

[fmassa]

Contributions are welcome!
Ideally, we would like the models to be trained on PyTorch using the examples/imagenet script.

[setuc]

We are working on developing the models for pytorch. we will make a pull request when done. Is there a need for pre-trained weights as well for such networks? if so which datasets? CIFAR-10? CIFAR-100?

[bnu-wangxun]

It's very kind of you to do such work. I think the pretrained weights on imagenet is needed,too. @setuc

[HaleTom]

This may be partially addressed by Add inceptionv4 pretrained model https://github.com/pytorch/vision/pull/43/