how can I run it with image files in folders ? such as imagenet

[younfor]

No description provided.

[MasoodK]

Not sure I understand what the TensorSpark-specific issue is here. That would be a standard pre-processing that you would need to do before feeding the dataset into the model. One way would be to write code (Python, Python with PySpark API, etc.) to read those images from those folders and save them all in one big CSV file, like the existing MNIST example. There might be other better ways as well :)

[illuzen]

Agree with @MasoodK here. That being said, I'm sure we would pull in useful, well-written pre-processing code that did this.