Andrea Palazzi

Hi @Xcrid , I can't really answer as I'm not much into pytorch data parallelism and lately I'm working in TF I'm keeping the issue open for someone else to...

I indeed tried to capture the same scene from multiple cameras for [this](https://arxiv.org/pdf/1706.08442.pdf) work. Still, I didn't find a clean way to do so. The workaround I used is to...

@mgarbade Hi Martin, at the end did you find a way to rotate the 3D scene s.t. it aligns to the RGB and depth image? Thanks

Thanks for your prompt response! I'll give it a try!

Hi @zabhishekgupta , you can implement the loader for your dataset in [`data_load.py`](https://github.com/ndrplz/transforming-autoencoders/blob/master/transforming_autoencoders/utils/data_load.py), then pass as argument the name of your new dataset in `main.py` with `--dataset` option

For MNIST dataset this takes place [here](https://github.com/ndrplz/transforming-autoencoders/blob/1ed0a4757c1abb21744fffc6235d761a2472f966/transforming_autoencoders/utils/data_load.py#L32-L36), which in turn calls [this](https://github.com/ndrplz/transforming-autoencoders/blob/1ed0a4757c1abb21744fffc6235d761a2472f966/transforming_autoencoders/utils/data_transform.py#L31-L88) function to transform the images. This returns for each split a list of [`TransformingAutoencoderExample`](https://github.com/ndrplz/transforming-autoencoders/blob/master/transforming_autoencoders/utils/data_structures.py). Each `TransformingAutoencoderExample` in turn...

Thanks @hughiephan. Yes, that is indeed the same dataset.