Removed most of readme for now.

The setup section is no longer needed because train scripts will
automatically create the data directory.

The FID section was misleading because FID computation is not currently
implemented during training.

The models section did not provide much useful information because there
is currently only one model.

README.md

@@ -1,42 +1,6 @@
 # Neural Terrain Generation
 Neural Terrain Generation (NTG) is a collection of generative neural networks that output heightmaps for 3D terrain. This repository contains code for the entire NTG training pipeline.
 
-## Setup
-After cloning, create a directory at the root of the repository called ``data``.
-
-## Computing FID Stats
-In order to compute FID stats, you'll need pretrained InceptionV3 weights. You can get these [here](https://huggingface.co/hayden-donnelly/inception-v3-fid/tree/main)
-in the form of ``inception_v3_fid.pickle``. Once you've downloaded this file, place it in the ``data`` directory at the root of the repository.
-
-If you would like to track FID as a model trains, you'll need to pre-compute the FID stats for the target dataset.
-This can be done by running the following command:
-```
-python fid --precompute --img_dir <PATH_TO_DATASET> --out_dir <PATH_TO_OUTPUT_DIRECTORY> --img_size <WIDTH> <HEIGHT>
-```
-Here's a specific example of the command:
-```
-python fid --precompute --img_dir ../heightmaps/world-heightmaps-01 --out_dir data/dataset_info --img_size 256 256
-```
-Outside of training, the FID of two datasets can be computed by specifying the path1 and path2 arguments instead of the img_dir argument:
-```
-python fid --path1 <PATH_TO_DATASET1> --path2 <PATH_TO_DATASET2> --out_dir <PATH_TO_OUTPUT_DIRECTORY> --img_size <WIDTH> <HEIGHT>
-```
-
-Full list of parameters:
-- ``--path1``: Path to image directory or .npz file containing pre-computed statistics. Default: ``None``
-- ``--path2``: Path to image directory or .npz file containing pre-computed statistics. Default: ``None``
-- ``--batch_size``: Batch size per device for computing the Inception activations. Default: ``50``
-- ``--img_size``: Resize images to this size. The format is (height, width). Default: ``None``, ``None``
-- ``--precompute``: If True, pre-compute statistics for given image directory. Default: ``False``
-- ``--img_dir``: Path to image directory for pre-computing statistics. Default: ``None``
-- ``--out_dir``: Path where pre-computed statistics are stored. Default: ``None``
-- ``--out_name``: Name of outputted statistics file. Default: ``stats``
-- ``--mmap``: If True, use mmap to compute statistics. Helpful for large datasets. Default: ``True``
-- ``--mmap_filename``: Name for mmap file. Only used if mmap is True. Default: ``data/temp/mmap_file``
-
-## Models
-- [Terra](./models/terra.py): a purely convolutional diffusion model.
-
 ## Development Environment
 The NTG development environment is managed with Nix. You can follow the steps below to get started.
 1. Install Nix with the [official installer](https://nixos.org/download/) or the [determinate installer](https://github.com/DeterminateSystems/nix-installer).