Hugging Face's logo

Datasets:
initiacms
/
OpticalRS-4M

Tasks:
Image Classification
ArXiv:
License:
Dataset card Files
xet
 Community
1
initiacms commited on
Commit
ebd44ab
·
verified ·
1 Parent(s): c654402

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +161 -0
README.md ADDED
@@ -0,0 +1,161 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ task_categories:
3
+ - image-classification
4
+ license: apache-2.0
5
+ ---
6
+
7
+ <p align="center">
8
+
9
+ <h2 align="center"><strong> Harnessing Massive Satellite Imagery with Efficient Masked Image Modeling</strong></h2>
10
+
11
+ <p align="center">
12
+ Fengxiang Wang<sup>1</sup>&nbsp;&nbsp;&nbsp;
13
+ Hongzhen Wang<sup>2,‡</sup>&nbsp;&nbsp;&nbsp;
14
+ Di Wang<sup>3</sup>&nbsp;&nbsp;&nbsp;
15
+ Zonghao Guo<sup>2</sup></br>
16
+ Zhenyu Zhong<sup>4</sup>&nbsp;&nbsp;&nbsp;
17
+ Long Lan<sup>1,‡</sup>&nbsp;&nbsp;
18
+ Wenjing Yang<sup>1,‡</sup>&nbsp;&nbsp;&nbsp;
19
+ Jing Zhang<sup>3,‡</sup>&nbsp
20
+ </br>
21
+ <sup>1</sup> National University of Defense Technology&nbsp;&nbsp;&nbsp;
22
+ <sup>2</sup>Tsinghua University &nbsp;&nbsp;&nbsp;</br>
23
+ <sup>3</sup>Wuhan University&nbsp;&nbsp;
24
+ <sup>4</sup>Nankai University&nbsp;&nbsp;
25
+ <div align='center' style="font-size: larger; "><strong>ICCV 2025</strong></div>
26
+
27
+ <p align="center">
28
+ 📃 <a href="https://arxiv.org/abs/2406.11933" target="_blank">Paper</a> |
29
+ 🤗 <a href="https://huggingface.co/datasets/initiacms/OpticalRS-4M" target="_blank"> OpticalRS-4M</a> |
30
+ 🤗 <a href="https://huggingface.co/datasets/initiacms/OpticalRS-13M" target="_blank"> OpticalRS-13M</a> |
31
+ 🤗 <a href="https://huggingface.co/initiacms/SelectiveMAE" target="_blank">Models</a>
32
+ </p>
33
+ </p>
34
+
35
+ ## 🎯Intruduction
36
+ - `Dataset`: `OpticalRS-13M` is a large-scale remote sensing dataset. This dataset, comprising 13 million optical images,
37
+ is designed to fully leverage the representation learning capabilities of MIM methods in RS applications, distinguished by its diverse scene details. We also offer a light version, named `OpticalRS-4M`.</br>
38
+ - `SelectiveMAE`: A novel and efficient MIM method tailored for remote sensing images. This method incorporates a new PSTS module,
39
+ which significantly accelerates convergence and enhances representation learning compared to the original MIM approach.
40
+
41
+
42
+ ## ✅ To do List
43
+ - [x] Initial release of checkpoint of SelectiveMAE.
44
+ - [x] Pretraining codes and configs for SelectiveMAE have be released.
45
+ - [x] OpticalRS-4M dataset has be released.
46
+ - [x] OpticalRS-13M dataset will be released.
47
+ - [ ] Codes and configs for downstream tasks of Scene Classification.
48
+ - [ ] Codes and configs for downstream tasks of Object Detection and Semantic Segmentation.
49
+
50
+
51
+
52
+ ## 🔥 News
53
+ - \[2025.06\] - **SelectiveMAE has been accepted by ICCV2025.**
54
+ - \[2025.06\] - OpticalRS-13M has been released on 🤗[HuggingFace](https://huggingface.co/datasets/initiacms/OpticalRS-13M).
55
+ - \[2025.06\] - Models have been released on 🤗[HuggingFace](https://huggingface.co/initiacms/SelectiveMAE).
56
+ - \[2025.06\] - OpticalRS-4M has been released on 🤗[HuggingFace](https://huggingface.co/datasets/initiacms/OpticalRS-4M).
57
+ - \[2025.06\] - The pretraining codes of the SelectiveMAE have been released.
58
+ - \[2024.06\] - Paper has been released on [arxiv](https://arxiv.org/abs/2406.11933).
59
+ - \[2024.06\] - The training logs and checkpoints of the SelectiveMAE have been released.
60
+
61
+
62
+
63
+ ## 📚 Contents
64
+
65
+ - [OpticalRS-4M](#opticalrs-4m)
66
+ - [OpticalRS-13M](#opticalrs-13m)
67
+ - [SelectiveMAE](#selectivemae)
68
+ - [Citation](#citation)
69
+ - [License](#license)
70
+
71
+
72
+ ## 🚀OpticalRS-4M
73
+ ### Usage
74
+ `OpticalRS-4M` available on 🤗HuggingFace via [OpticalRS-4M](https://huggingface.co/datasets/initiacms/OpticalRS-4M).
75
+
76
+ Use the following command to unzip:
77
+ ```bash
78
+ # if 7z is available
79
+ 7z x OpticalRS-4M.zip
80
+ # if zip and unzip is available
81
+ zip -s 0 OpticalRS-4M.zip --out whole.zip
82
+ unzip whole.zip
83
+ ```
84
+
85
+ ### Experiments on OpticalRS-4M
86
+ OpticalRS-4M offers a significantly larger and more diverse image set compared to previous datasets. To evaluate its effectiveness, we pre-train a **ViT-Base** model using the vanilla **MAE** method.
87
+ For comparison, we use the [**MillionAID**](https://captain-whu.github.io/DiRS/) dataset, maintaining an equal number of data points during training: 800 epochs for **MillionAID**'s 1 million images and 200 epochs for our **OpticalRS-4M** dataset.
88
+
89
+ | Dataset | Pretrained model | Images Number | Epoch | Sence Classification | Sence Classification | Object Detection | Object Detection | Semantic Segmentation | Semantic Segmentation|
90
+ |:----------:|:----------------:|:-------------:|:-----:|:---------------------:|:---------------------------:|:-------------------------:|:-----------------:|:--------:|:------------:|
91
+ | | | | | AID | RESISC-45 | DIOR | DIOR-R | LoveDA | SpaceNetv1 |
92
+ | | | | | OA (TR=20%/50%) | OA (TR=20%/50%) | mAP50 | mAP50 | mIoU | mF1 |
93
+ | MillionAID | [Weights](https://pan.baidu.com/s/1OCl7whWnYoyrAI8zha_Kbg?pwd=0330) | 1 million | 800 | 94.92/97.38 | 89.20/93.60 | 71.80 | 62.33 | 51.24 | 79.24 |
94
+ | OpticalRS-4M | [Weights](https://pan.baidu.com/s/1-6HBRbAyHMUrTSwcSOIhyw?pwd=0330) | 2 million | 400 | 96.64/98.10 | 91.80/94.31 | 73.90 | 65.95 | 52.86 | 79.37 |
95
+ | OpticalRS-4M | [Weights](https://pan.baidu.com/s/1S_oTibDouAi-VrmESn7qPg?pwd=0330) | 3 million | 267 | 96.67/98.18 | 92.24/94.41 | 75.40 | 67.07 | 52.39 | 79.37 |
96
+ | OpticalRS-4M | [Weights](https://pan.baidu.com/s/1zmS24CqFo44Rkkkl2YqeaQ?pwd=0330) | 4 million | 200 | 96.10/98.03 | 92.38/94.30 | 74.70 | 66.26 | 52.75 | 79.23 |
97
+ | OpticalRS-4M | [Weights](https://pan.baidu.com/s/1Qrgtv7Dotfb_QQ2GCk6bog?pwd=0330) | 4 million | 800 | **96.88/98.22** | **92.44/94.43** | **75.40** | **67.35** | **52.80** | **79.41** |
98
+
99
+
100
+ ## 🚀OpticalRS-13M
101
+ `OpticalRS-13M` available on 🤗HuggingFace via [OpticalRS-13M](https://huggingface.co/datasets/initiacms/OpticalRS-13M). Follow OpticalRS-4M to unzip.
102
+
103
+ ## 🚀SelectiveMAE
104
+
105
+ ### :gear: Installation for Pretraining
106
+ Please install the pretraining dependencies in `SelectiveMAE/requirements.txt`:
107
+ ```sh
108
+ # Optionally create a conda environment
109
+ conda create -n selectivemae python=3.10 -y
110
+ conda activate selectivemae
111
+ # Install dependencies
112
+ pip install -r requirements.txt
113
+ ```
114
+
115
+ ### :blue_car: Pretraining for SelectiveMAE
116
+
117
+ To pre-train ViT-Base, run the following on 8 GPUs:
118
+ ```sh
119
+ torchrun --nproc_per_node=8 --nnodes 1 --master_port 16666 main_pretrain.py --batch_size 256 --selectivemae --dataset opticalrs-4m --dataset_path 'your_dataset_path' --model mae_vit_base_patch16 --output_dir output --norm_pix_loss --blr 1.5e-4 --weight_decay 0.05 --num_workers 12 --decoder_depth 12 --mask_ratio 0.85 --kept_mask_ratio 0.25 --epochs 800 --warmup_epochs 30
120
+ ```
121
+ First, download the corresponding dataset, then set `opticalrs-4m` or `opticalrs-13m`, and update the dataset path accordingly. To train ViT-Small or ViT-Large, set `--model mae_vit_small_patch16` or `--model mae_vit_large_patch16`. You can use `--accum_iter` to perform gradient accumulation if your hardware could not fit the batch size. [FlashAttention 2](https://github.com/Dao-AILab/flash-attention) should be installed with `pip install flash-attn --no-build-isolation`.
122
+
123
+
124
+ ### :rocket: Results on downstream tasks
125
+
126
+ | Model | Publication | Backbone | Sence Classification | Sence Classification | Object Detection | Object Detection | Semantic Segmentation | Semantic Segmentation |
127
+ |--------------|:-----------:|:----------:|:---------------------:|:-----------------:|:----------:|:----------:|:------------:|:----------:|
128
+ | | | | AID | RESISC-45 | DIOR | DIOR-R | LoveDA | SpaceNetv1 |
129
+ | | | | OA (TR=20%/50%) | OA (TR=20%/50%) | mAP50 | mAP50 | mIoU | mF1 |
130
+ | SeCo | ICCV'21 | ResNet-50 | 93.47/95.99 | 89.64/92.91 | - | - | 43.63 | 77.09 |
131
+ | GASSL | ICCV'21 | ResNet-50 | 93.55/95.92 | 90.86/93.06 | 67.40 | 65.65 | 48.76 | 78.51 |
132
+ | TOV | JSTARS'23 | ResNet-50 | 95.16/97.09 | 90.97/93.79 | 70.16 | 66.33 | 49.70 | - |
133
+ | CACo | CVPR'23 | ResNet-50 | 90.88/95.05 | 88.28/91.94 | 66.91 | 64.10 | 48.89 | 77.94 |
134
+ | SatMAE | NIPS'22 | ViT-L | 95.02/96.94 | 91.72/94.10 | 70.89 | 65.66 | - | 78.07 |
135
+ | ScaleMAE | ICCV'23 | ViT-L | 96.44/97.58 | 92.63/95.04 | 73.81 | 66.47 | - | - |
136
+ | SSL4EO | GRSM'23 | ViT-S | 91.06/94.74 | 87.60/91.27 | 64.82 | 61.23 | - | - |
137
+ | RingMo | TGRS'22 | Swin-B | 96.90/98.34 | 94.25/95.67 | 75.90 | - | - | - |
138
+ | SatLas | ICCV'23 | Swin-B | 94.96/97.38 | 92.16/94.70 | 74.10 | 67.59 | - | - |
139
+ | GFM | ICCV'23 | Swin-B | 95.47/97.09 | 92.73/94.64 | 72.84 | 67.67 | - | - |
140
+ | RVSA | TGRS'23 | ViT-B+RVSA | 97.03/98.50 | 93.93/95.69 | 75.80 | 68.06 | 51.95 | - |
141
+ | SelectiveMAE(OpticalRS-4M) | [Baidu](https://pan.baidu.com/s/1Y4WBj35-HAKeZJe125TG8Q?pwd=0330) & [HuggingFace](https://huggingface.co/initiacms/SelectiveMAE) | ViT-B | 96.90/98.12 | 93.35/94.58 | 75.70 | 67.78 | 53.05 | **79.50** |
142
+ | SelectiveMAE(OpticalRS-4M)| [Baidu](https://pan.baidu.com/s/1miSlmoeZLjzc_WgXE87Fxg?pwd=0330) & [HuggingFace](https://huggingface.co/initiacms/SelectiveMAE) | ViT-L | 97.25/98.48 | 94.57/95.77 | 77.80 | 70.31 | **54.31** | 79.46 |
143
+ | SelectiveMAE(OpticalRS-13M) | [Baidu](https://pan.baidu.com/s/1_QNLBGhrViapquDcVZHnkw?pwd=bmzj) & [HuggingFace](https://huggingface.co/initiacms/SelectiveMAE) | ViT-B | 97.10/98.28 | 93.70/95.48 | 75.80 | 67.69 | 52.68 | 79.44 |
144
+ | SelectiveMAE(OpticalRS-13M)| [Baidu](https://pan.baidu.com/s/10HJ_kZwW2nxNqDNjJRb6SQ?pwd=eyjn) & [HuggingFace](https://huggingface.co/initiacms/SelectiveMAE) | ViT-L | **97.49/98.52** | **94.73/96.36** | **78.70** | **71.75** | 53.92 | 79.48 |
145
+
146
+ # 🔗Citation
147
+ If you find SelectiveMAE helpful, please consider citing:
148
+
149
+ ```latex
150
+ @article{selectivemae,
151
+ title={Harnessing Massive Satellite Imagery with Efficient Masked Image Modeling},
152
+ author={Fengxiang Wang and Hongzhen Wang and Di Wang and Zonghao Guo and Zhenyu Zhong and Long Lan and Wenjing Yang and Jing Zhang},
153
+ year={2025},
154
+ journal={arXiv preprint arXiv:2406.11933},
155
+ }
156
+
157
+ ```
158
+
159
+ ## 🤝License
160
+
161
+ This work is under the [Apache License Version 2.0](https://www.apache.org/licenses/LICENSE-2.0), while some specific operations in this codebase might be with other licenses. Please refer to [LICENSE.md](docs/LICENSE.md) for a more careful check, if you are using our code for commercial matters.