https://github.com/mrbid
https://huggingface.co/tfnn
https://www.kaggle.com/mrbiid

---

In this model I train 32^3 (32768) networks each producing the grayscale value of one voxel, this means there are 32,768 train_y targets.

You just execute `multi_train_silent.sh` or `multi_train.sh` to train all 32^3 networks, once it is done you call `pred_multi.sh` to create the prediction volume that outputs to the `pred_multi/` directory, then you call `cat_pred_multi.sh` to concatenate all the predictions into a single .csv and a single .dat file. You can visualise this prediction from the CSV using `python3 view_pred_final.py`.

You can create a new `pred_input.npy` by calling `python3 pred.py imagepath.jpg` this allows you to feed custom images into the model for the prdiction process.


