search

ViTAE-Transformer / ViTAE-Transformer-Remote-Sensing

A comprehensive list [SAMRS@NeurIPS'23, RVSA@TGRS'22, RSP@TGRS'22] of our research works related to remote sensing, including papers, codes, and citations. Note: The repo for [TGRS'22] "An Empirical Study of Remote Sensing Pretraining" has been moved to: https://github.com/ViTAE-Transformer/RSP
462 stars 53 forks source link

What does this ‘BIT’ abbreviation stand for #6

Closed zgsxwsdxg closed 2 years ago

zgsxwsdxg commented 2 years ago

hello，great job. What does this ‘BIT’ abbreviation stand for? ths.

DotWang commented 2 years ago

'BIT' is a change detection method, it is proposed by the paper:

H. Chen, Z. Qi, and Z. Shi, “Remote Sensing Image Change Detection With Transformers,” IEEE Trans. Geosci. Remote Sens., vol. 60, p.3095166, Jan. 2022

zgsxwsdxg commented 2 years ago

'BIT' is a change detection method, it is proposed by the paper:

H. Chen, Z. Qi, and Z. Shi, “Remote Sensing Image Change Detection With Transformers,” IEEE Trans. Geosci. Remote Sens., vol. 60, p.3095166, Jan. 2022

Ths，How to understand this sentence ‘For LEVIR dataset, clip the images to 256 × 256 patches and change the structure to’？Whether the aspect provides scripts to process the data？I download levir dataset，but the dataset is 1024 and There is no output folder, just a label folder.

DotWang commented 2 years ago

'BIT' is a change detection method, it is proposed by the paper: H. Chen, Z. Qi, and Z. Shi, “Remote Sensing Image Change Detection With Transformers,” IEEE Trans. Geosci. Remote Sens., vol. 60, p.3095166, Jan. 2022

Ths，How to understand this sentence ‘For LEVIR dataset, clip the images to 256 × 256 patches and change the structure to’？Whether the aspect provides scripts to process the data？I download levir dataset，but the dataset is 1024 and There is no output folder, just a label folder.

You need to crop the images, labels and organize the folders, it is easy to implement.

zgsxwsdxg commented 2 years ago

'BIT' is a change detection method, it is proposed by the paper: H. Chen, Z. Qi, and Z. Shi, “Remote Sensing Image Change Detection With Transformers,” IEEE Trans. Geosci. Remote Sens., vol. 60, p.3095166, Jan. 2022

Ths，How to understand this sentence ‘For LEVIR dataset, clip the images to 256 × 256 patches and change the structure to’？Whether the aspect provides scripts to process the data？I download levir dataset，but the dataset is 1024 and There is no output folder, just a label folder.

You need to crop the images, labels and organize the folders, it is easy to implement.

我的英文不太好，还是中文问下你吧。 就是裁切确实容易实现，我不太明白的是：如果随机裁切，那么每个人的样本不一样，训练的模型MAP和paper是否有大的gap，如果有大的gap，怎么确定问题在哪？怎么能排除不是裁切的问题？你是随机裁切的么？谢谢

DotWang commented 2 years ago

你说的随机裁剪是啥意思，这个数据集是1024*1024的，讲道理你裁剪成256*256是不用设置overlap的，而且这个数据集训练集验证集和测试集是固定的啊，不存在什么随机性吧

DotWang commented 2 years ago

我貌似懂了，你的意思是训练过程中在数据增强部分随机裁剪？那貌似不太行，这个LEVIR数据集应该是先裁剪，变成图像尺寸为256*256的新数据集，再用这个整理好的新数据集拿去训练

zgsxwsdxg commented 2 years ago

我貌似懂了，你的意思是训练过程中在数据增强部分随机裁剪？那貌似不太行，这个LEVIR数据集应该是先裁剪，变成图像尺寸为256*256的新数据集，再用这个整理好的新数据集拿去训练

谢谢你的耐心解答，我知道了，你的裁切就是等分裁切成256x256，这样的话样本的确定的。再次感谢，如果有不懂的还得请教你。