Skip to content

Commit fcfded5

Browse files
authored
Update README.md
1 parent 2795a72 commit fcfded5

File tree

1 file changed

+28
-3
lines changed

1 file changed

+28
-3
lines changed

README.md

+28-3
Original file line numberDiff line numberDiff line change
@@ -1,13 +1,36 @@
11
# Scale-MAE 🛰️
22

3-
## Requirements and Setup ##
4-
Set up your Python (`conda` suggested) environment in the same way as the [MAE](github.com/facebookresearch/mae) repository. Please note that the `timm==0.3.2` patch is still needed for Scale-MAE.
3+
![image](https://user-images.githubusercontent.com/1455579/217665789-b46d6830-445f-4151-b7a4-a2152a81a8d1.png)
4+
5+
6+
This repository provides a reimplementation of the code for [Scale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learning](https://arxiv.org/abs/2212.14532) (the original code was optimized for our distributed cluster).
7+
8+
```
9+
@article{reed2022scale,
10+
title={Scale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learning},
11+
author={Reed, Colorado J and Gupta, Ritwik and Li, Shufan and Brockman, Sarah and Funk, Christopher and Clipp, Brian and Candido, Salvatore and Uyttendaele, Matt and Darrell, Trevor},
12+
journal={arXiv preprint arXiv:2212.14532},
13+
year={2022}
14+
}
15+
```
16+
17+
* This repo is a modification on the [MAE repo](https://github.com/facebookresearch/mae). Installation and preparation follow that repo ;-).
18+
19+
* As mentioned in the MAE repo, this repo is based on [`timm==0.3.2`](https://github.com/rwightman/pytorch-image-models), for which a [fix](https://github.com/rwightman/pytorch-image-models/issues/420#issuecomment-776459842) is needed to work with PyTorch 1.8.1+. In addition, install gdal, rasterio, and Shapely. This tends to work pretty well (but gdal is notoriously tricky):
20+
21+
```bash
22+
conda install geopandas. # this should setup gdal correctly...
23+
pip install rasterio shapely
24+
```
525

6-
In addition, install `GDAL`, `rasterio`, and `Shapely`.
726

827
## Pretraining ##
928
Datasets are defined by config files in `config`
1029

30+
31+
32+
# TOD FIX ALL OF THIS
33+
1134
```
1235
python -m torch.distributed.launch --nproc_per_node=4 --master_port=11083 main_pretrain.py\
1336
--batch_size 256 \
@@ -70,4 +93,6 @@ main_linprobe.py \
7093
flag `--finetune` is enabled for fine-tuning and not enabled for linear probing
7194

7295

96+
---
97+
7398
> Note: THIS SOFTWARE AND/OR DATA WAS DEPOSITED IN THE BAIR OPEN RESEARCH COMMONS REPOSITORY ON 2/8/23.

0 commit comments

Comments
 (0)