作者:AI小浩@公众号
编辑:3D视觉开发者社区
论文地址:https://arxiv.org/pdf/2108.10257.pdf
训练代码下载:https://github.com/cszn/KAIR
测试:https://github.com/JingyunLiang/SwinIR
预训练模型下载:
https://github.com/JingyunLiang/SwinIR/releases
论文翻译:https://blog.csdn.net/hhhhhhhhhhwwwwwwwwww/article/details/124434886
本文介绍如何使用SwinIR和预训练模型实现图像的超分。
下载代码:
https://github.com/JingyunLiang/SwinIR,然后解压,如下图:
然后下载预训练权重文件,打开连接:https://github.com/JingyunLiang/SwinIR/releases,如下图:
预训练权重有很多,选择自己用到的预训练权重下载,下载后将其放入./model_zoo/swinir文件夹下面,如下图:
然后执行命令:
python main_test_swinir.py --task classical_sr --scale 2 --training_patch_size 48 --model_path model_zoo/swinir/001_classicalSR_DIV2K_s48w8_SwinIR-M_x2.pth --folder_lq testsets/Set5/LR_bicubic/X2 --folder_gt testsets/Set5/HR
task:任务类型,包括:classical_sr, lightweight_sr, real_sr, gray_dn, color_dn, jpeg_car。
scale:放大的倍数,这个和预训练权重对应。比如这次设置的是2,对应权重的x2。
training_patch_size:用于训练 SwinIR 的patch大小。
model_path:预训练模型的路径。
folder_lq:低质量图片的文件夹路径。
folder_gt:高质量图片的文件夹路径。
更多的命令如下:
# 001 Classical Image Super-Resolution (middle size)
# Note that --training_patch_size is just used to differentiate two different settings in Table 2 of the paper. Images are NOT tested patch by patch.
# (setting1: when model is trained on DIV2K and with training_patch_size=48)
python main_test_swinir.py --task classical_sr --scale 2 --training_patch_size 48 --model_path model_zoo/swinir/001_classicalSR_DIV2K_s48w8_SwinIR-M_x2.pth --folder_lq testsets/Set5/LR_bicubic/X2 --folder_gt testsets/Set5/HR
python main_test_swinir.py --task classical_sr --scale 3 --training_patch_size 48 --model_path model_zoo/swinir/001_classicalSR_DIV2K_s48w8_SwinIR-M_x3.pth --folder_lq testsets/Set5/LR_bicubic/X3 --folder_gt testsets/Set5/HR
python main_test_swinir.py --task classical_sr --scale 4 --training_patch_size 48 --model_path model_zoo/swinir/001_classicalSR_DIV2K_s48w8_SwinIR-M_x4.pth --folder_lq testsets/Set5/LR_bicubic/X4 --folder_gt testsets/Set5/HR
python main_test_swinir.py --task classical_sr --scale 8 --training_patch_size 48 --model_path model_zoo/swinir/001_classicalSR_DIV2K_s48w8_SwinIR-M_x8.pth --folder_lq testsets/Set5/LR_bicubic/X8 --folder_gt testsets/Set5/HR
# (setting2: when model is trained on DIV2K+Flickr2K and with training_patch_size=64)
python main_test_swinir.py --task classical_sr --scale 2 --training_patch_size 64 --model_path model_zoo/swinir/001_classicalSR_DF2K_s64w8_SwinIR-M_x2.pth --folder_lq testsets/Set5/LR_bicubic/X2 --folder_gt testsets/Set5/HR
python main_test_swinir.py --task classical_sr --scale 3 --training_patch_size 64 --model_path model_zoo/swinir/001_classicalSR_DF2K_s64w8_SwinIR-M_x3.pth --folder_lq testsets/Set5/LR_bicubic/X3 --folder_gt testsets/Set5/HR
python main_test_swinir.py --task classical_sr --scale 4 --training_patch_size 64 --model_path model_zoo/swinir/001_classicalSR_DF2K_s64w8_SwinIR-M_x4.pth --folder_lq testsets/Set5/LR_bicubic/X4 --folder_gt testsets/Set5/HR
python main_test_swinir.py --task classical_sr --scale 8 --training_patch_size 64 --model_path model_zoo/swinir/001_classicalSR_DF2K_s64w8_SwinIR-M_x8.pth --folder_lq testsets/Set5/LR_bicubic/X8 --folder_gt testsets/Set5/HR
# 002 Lightweight Image Super-Resolution (small size)
python main_test_swinir.py --task lightweight_sr --scale 2 --model_path model_zoo/swinir/002_lightweightSR_DIV2K_s64w8_SwinIR-S_x2.pth --folder_lq testsets/Set5/LR_bicubic/X2 --folder_gt testsets/Set5/HR
python main_test_swinir.py --task lightweight_sr --scale 3 --model_path model_zoo/swinir/002_lightweightSR_DIV2K_s64w8_SwinIR-S_x3.pth --folder_lq testsets/Set5/LR_bicubic/X3 --folder_gt testsets/Set5/HR
python main_test_swinir.py --task lightweight_sr --scale 4 --model_path model_zoo/swinir/002_lightweightSR_DIV2K_s64w8_SwinIR-S_x4.pth --folder_lq testsets/Set5/LR_bicubic/X4 --folder_gt testsets/Set5/HR
# 003 Real-World Image Super-Resolution (use --tile 400 if you run out-of-memory)
# (middle size)
python main_test_swinir.py --task real_sr --scale 4 --model_path model_zoo/swinir/003_realSR_BSRGAN_DFO_s64w8_SwinIR-M_x4_GAN.pth --folder_lq testsets/RealSRSet+5images --tile
# (larger size + trained on more datasets)
python main_test_swinir.py --task real_sr --scale 4 --large_model --model_path model_zoo/swinir/003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN.pth --folder_lq testsets/RealSRSet+5images
# 004 Grayscale Image Deoising (middle size)
python main_test_swinir.py --task gray_dn --noise 15 --model_path model_zoo/swinir/004_grayDN_DFWB_s128w8_SwinIR-M_noise15.pth --folder_gt testsets/Set12
python main_test_swinir.py --task gray_dn --noise 25 --model_path model_zoo/swinir/004_grayDN_DFWB_s128w8_SwinIR-M_noise25.pth --folder_gt testsets/Set12
python main_test_swinir.py --task gray_dn --noise 50 --model_path model_zoo/swinir/004_grayDN_DFWB_s128w8_SwinIR-M_noise50.pth --folder_gt testsets/Set12
# 005 Color Image Deoising (middle size)
python main_test_swinir.py --task color_dn --noise 15 --model_path model_zoo/swinir/005_colorDN_DFWB_s128w8_SwinIR-M_noise15.pth --folder_gt testsets/McMaster
python main_test_swinir.py --task color_dn --noise 25 --model_path model_zoo/swinir/005_colorDN_DFWB_s128w8_SwinIR-M_noise25.pth --folder_gt testsets/McMaster
python main_test_swinir.py --task color_dn --noise 50 --model_path model_zoo/swinir/005_colorDN_DFWB_s128w8_SwinIR-M_noise50.pth --folder_gt testsets/McMaster
# 006 JPEG Compression Artifact Reduction (middle size, using window_size=7 because JPEG encoding uses 8x8 blocks)
python main_test_swinir.py --task jpeg_car --jpeg 10 --model_path model_zoo/swinir/006_CAR_DFWB_s126w7_SwinIR-M_jpeg10.pth --folder_gt testsets/classic5
python main_test_swinir.py --task jpeg_car --jpeg 20 --model_path model_zoo/swinir/006_CAR_DFWB_s126w7_SwinIR-M_jpeg20.pth --folder_gt testsets/classic5
python main_test_swinir.py --task jpeg_car --jpeg 30 --model_path model_zoo/swinir/006_CAR_DFWB_s126w7_SwinIR-M_jpeg30.pth --folder_gt testsets/classic5
python main_test_swinir.py --task jpeg_car --jpeg 40 --model_path model_zoo/swinir/006_CAR_DFWB_s126w7_SwinIR-M_jpeg40.pth --folder_gt testsets/classic5
Result
那么如何用起来呢?
我们使用classical_sr任务讲解。
首先将评分的部分注释掉。注释main_test_swinir.py中的83行到119行。
将185行的folder_gt改为folder_lq
修改218行到223行的代码。由
# 001 classical image sr/ 002 lightweight image sr (load lq-gt image pairs)
if args.task in ['classical_sr', 'lightweight_sr']:
img_gt = cv2.imread(path, cv2.IMREAD_COLOR).astype(np.float32) / 255.
img_lq = cv2.imread(f'{args.folder_lq}/{imgname}x{args.scale}{imgext}', cv2.IMREAD_COLOR).astype(
np.float32) / 255.
修改为:
# 001 classical image sr/ 002 lightweight image sr (load lq-gt image pairs)
if args.task in ['classical_sr', 'lightweight_sr']:
# img_gt = cv2.imread(path, cv2.IMREAD_COLOR).astype(np.float32) / 255.
img_gt = cv2.imdecode(np.fromfile(path, dtype=np.uint8), -1)
img_gt = cv2.cvtColor(img_gt, cv2.IMREAD_COLOR).astype(np.float32) / 255.
lq_path=f'{args.folder_lq}/{imgname}{imgext}'
img_lq = cv2.imdecode(np.fromfile(lq_path, dtype=np.uint8), -1)
img_lq = cv2.cvtColor(img_lq, cv2.IMREAD_COLOR).astype(np.float32) / 255.
使用cv2.imdecode方法读取图片,可以解决路径有中文的问题。
将81行的代码
cv2.imwrite(f'{save_dir}/{imgname}_SwinIR.png', output)
修改为:
write_path=f'{save_dir}/{imgname}.jpg'
cv2.imencode('.jpg', output)[1].tofile(write_path) # 保存图片
这样,生成的图片和原来低质量的图片名字保持一致了。
完成上面的修改,执行命令:
python main_test_swinir.py --task classical_sr --scale 2 --training_patch_size 64 --model_path model_zoo/swinir/001_classicalsr_df2k_s64w8_swinir-m_x2.pth --folder_lq testsets/Set5/LR_bicubic/X2
等待命令运行完成后,我们就在result文件夹下面看到超分后的图像了。
对比结果:
我在对这些图片超分后,然后执行分类任务测试。在实际的测试中,和超分的图片作对比有了0.1%的提升,提升成绩有限。
我推测:超分后,只是让我们人眼觉得图像的画质提高了,但是对于神经网络来说,有用的特征并没有得到提升。图像在输入神经网络后,一般要经过Resize,超分前的图像尺寸接近224×224,超分后,提升了像素,但是输入进网络的时候又Resize成224×224,这样就导致了有效的特征被压缩了,所以提升有限。
完整代码:https://download.csdn.net/download/hhhhhhhhhhwwwwwwwwww/85251176
版权声明:本文为奥比中光3D视觉开发者社区特约作者授权原创发布,未经授权不得转载,本文仅做学术分享,版权归原作者所有,若涉及侵权内容请联系删文。
3D视觉开发者社区是由奥比中光给所有开发者打造的分享与交流平台,旨在将3D视觉技术开放给开发者。平台为开发者提供3D视觉领域免费课程、奥比中光独家资源与专业技术支持。点击加入3D视觉开发者社区,和开发者们一起讨论分享吧~
或可微信关注官方公众号 3D视觉开发者社区 ,获取更多干货知识哦。