## Motivation
When processing data in the isaid experiment, generated images only have
binary pixel values of 0 or 1 instead of the corresponding class values.
This causes significant interference and prevents subsequent experiments
from proceeding. After investigation, it was found that the issue was
caused by using the wrong image format for saving the images. Saving the
images as PNG resulted in binary pixel values, while saving the images
as BMP resolved the issue and correctly saved the class values.
## Modification
`img_patch.save(save_path_image, format='BMP')`
This code will save the image data as BMP format.
## BC-breaking
Confirm that the modification does not introduce new issues and test
that the modified code successfully resolves the original problem.
---------
Co-authored-by: 谢昕辰 <xiexinch@outlook.com>
Co-authored-by: CSH <40987381+csatsurnh@users.noreply.github.com>
## Motivation
The original version of Visual Attention Network (VAN) can be found from
https://github.com/Visual-Attention-Network/VAN-Segmentation
添加Visual Attention Network (VAN)的支持。
## Modification
added a floder mmsegmentation/projects/van/
added 13 configs totally and aligned performance basically.
只增加了一个文件夹，共增加13个配置文件，基本对齐性能（没有全部跑）。
## Use cases (Optional)
Before running, you may need to download the pretrain model from
https://cloud.tsinghua.edu.cn/d/0100f0cea37d41ba8d08/
and then move them to the folder mmsegmentation/pretrained/, i.e.
"mmsegmentation/pretrained/van_b2.pth".
After that, run the following command:
cd mmsegmentation
bash tools/dist_train.sh
projects/van/configs/van/van-b2_pre1k_upernet_4xb2-160k_ade20k-512x512.py
4
---------
Co-authored-by: xiexinch <xiexinch@outlook.com>
## Motivation
While customizing the number of samples using `ann_file` for Cityscapes,
I noticed that when the `ann_file` name is incorrect, it will silently
resort to loading the dataset from the directory.
I think when the user intends to load using `ann_file`, it should not
silently fail, but give some sort of error message or warning.
## Modification
I added assertion to check whether the `ann_file` exists instead of
silently resorting to loading from the directory.
Since `ann_file` is set to `''` by default and joined with
`self.data_root`, I used `osp.isdir` to first check if `self.ann_dir` is
a directory or text file.
## BC-breaking (Optional)
Not that I am aware of.
## Use cases (Optional)
If this PR introduces a new feature, it is better to list some use cases
here, and update the documentation.
---------
Co-authored-by: 谢昕辰 <xiexinch@outlook.com>
Co-authored-by: CSH <40987381+csatsurnh@users.noreply.github.com>
## Motivation
For support with reading multiple remote sensing image formats, please
refer to https://gdal.org/drivers/raster/index.html.
Byte, UInt16, Int16, UInt32, Int32, Float32, Float64, CInt16, CInt32,
CFloat32 and CFloat64 are supported for reading and writing.
Support input of two images for change detection tasks, and support the
LEVIR-CD dataset.
## Modification
Add LoadSingleRSImageFromFile in 'mmseg/datasets/transforms/loading.py'.
Load a single remote sensing image for object segmentation tasks.
Add LoadMultipleRSImageFromFile in
'mmseg/datasets/transforms/loading.py'.
Load two remote sensing images for change detection tasks.
Add ConcatCDInput in 'mmseg/datasets/transforms/transforms.py'.
Combine images that have been separately augmented for data enhancement.
Add BaseCDDataset in 'mmseg/datasets/basesegdataset.py'
Base class for datasets used in change detection tasks.
---------
Co-authored-by: xiexinch <xiexinch@outlook.com>
Motivation
Typo in docs/en/user_guides/visualization_feature_map.md.
Modification
reature -> feature
Checklist
- [x] Pre-commit or other linting tools are used to fix the potential
lint issues.
- [x] The modification is covered by complete unit tests. If not, please
add more unit test to ensure the correctness.
- [x] If the modification has potential influence on downstream
projects, this PR should be tested with downstream projects, like MMDet
or MMDet3D.
- [x] The documentation has been modified accordingly, like docstring or
example tutorials.
