swift-MetalColor iOS实时自动图像着色基于CNN和Metal技术

Swift 和 Metal 的结合可为 iOS 开发带来强大的图像能力,尤其是在计算密集型任务上。通过MetalCore ML,开发者能够直接在设备上实现高效的神经网络推理,像实时自动着色这样的任务变得比较简单。在swift-MetalColor项目中,借助卷积神经网络(CNN)模型,从灰度图像中推测出色彩,让图像自动着色。整个过程从图像预到加载 CNN 模型,再到通过 GPU 进行推理,在设备上实时显示,极大提升了图像的效率。使用Metal Performance Shaders优化图形性能,使得这一切可以在本地进行,无需依赖云端,提升了响应速度。对于需要实时图像的开发者来说,这个项目了一个好的技术参考。

zip
swift-MetalColor在iOS上使用CNN自动实时着色.zip 预估大小:93个文件
folder
MetalColor-master 文件夹
folder
MetalColor.xcodeproj 文件夹
folder
project.xcworkspace 文件夹
file
contents.xcworkspacedata 155B
file
project.pbxproj 45KB
file
result1.jpeg 62KB
folder
MetalColor 文件夹
file
ViewController.swift 5KB
folder
Assets.xcassets 文件夹
folder
AppIcon.appiconset 文件夹
file
Contents.json 1KB
folder
MetalNN 文件夹
file
ColorNet.swift 4KB
file
Convolution.swift 4KB
file
UpscaleAdd.swift 2KB
file
BatchNorm.swift 2KB
file
Shaders.metal 5KB
file
Helper.swift 3KB
file
IO.swift 3KB
folder
Base.lproj 文件夹
file
LaunchScreen.storyboard 2KB
file
Main.storyboard 5KB
file
AppDelegate.swift 2KB
file
Info.plist 1KB
folder
weights 文件夹
file
vgg16_conv2_2_biases 512B
file
uv_filter 216B
file
color3_color3_bn_beta 512B
file
color3_color3_bn_mean 512B
file
color4_color4_bn_beta 1024B
file
vgg16_conv4_3_filter 9MB
file
uv_uv_bn_mean 8B
file
pool4_bn_mean 2KB
file
pool1_bn_mean 256B
file
color2_color2_bn_gamma 256B
file
uv_uv_bn_gamma 8B
file
color1_color1_bn_variance 12B
file
color1_filter 7KB
file
vgg16_conv4_1_biases 2KB
file
vgg16_conv1_1_filter 7KB
file
color0_filter 324B
file
pool2_bn_mean 512B
file
color4_filter 512KB
file
color1_color1_bn_gamma 12B
file
pool1_bn_beta 256B
file
vgg16_conv4_2_biases 2KB
file
vgg16_conv3_3_biases 1024B
file
color3_filter 1.13MB
file
vgg16_conv4_3_biases 2KB
file
color3_biases 512B
file
color2_biases 256B
file
color2_color2_bn_beta 256B
file
pool4_bn_gamma 2KB
file
color0_color0_bn_mean 12B
file
color1_color1_bn_beta 12B
file
color0_color0_bn_variance 12B
file
pool2_bn_beta 512B
file
color3_color3_bn_gamma 512B
file
color0_color0_bn_gamma 12B
file
vgg16_conv3_3_filter 2.25MB
file
pool3_bn_mean 1024B
file
uv_biases 8B
file
vgg16_conv3_1_biases 1024B
file
color1_color1_bn_mean 12B
file
pool2_bn_gamma 512B
file
uv_uv_bn_beta 8B
file
pool3_bn_variance 1024B
file
vgg16_conv1_1_biases 256B
file
color2_color2_bn_variance 256B
file
vgg16_conv2_2_filter 576KB
file
vgg16_conv4_1_filter 4.5MB
file
pool1_bn_variance 256B
file
pool3_bn_gamma 1024B
file
vgg16_conv3_2_filter 2.25MB
file
color0_color0_bn_beta 12B
file
vgg16_conv3_2_biases 1024B
file
vgg16_conv2_1_biases 512B
file
vgg16_conv3_1_filter 1.13MB
file
pool1_bn_gamma 256B
file
uv_uv_bn_variance 8B
file
pool4_bn_beta 2KB
file
color0_biases 12B
file
color2_filter 288KB
file
pool3_bn_beta 1024B
file
color4_biases 1024B
file
vgg16_conv1_2_filter 144KB
file
vgg16_conv2_1_filter 288KB
file
color4_color4_bn_gamma 1024B
file
vgg16_conv1_2_biases 256B
file
color4_color4_bn_variance 1024B
file
pool2_bn_variance 512B
file
color3_color3_bn_variance 512B
file
vgg16_conv4_2_filter 9MB
file
pool4_bn_variance 2KB
file
color4_color4_bn_mean 1024B
file
color2_color2_bn_mean 256B
file
color1_biases 12B
file
test1.jpg 110KB
file
test3.jpg 217KB
file
result2.jpeg 53KB
file
README.md 509B
file
test2.jpeg 26KB
...
zip 文件大小:29.36MB