VGG：小型滤波器的堆叠
通过简单的 3x3 模块构建深度
VGG：小型滤波器的堆叠 是 CoddyKit 上的免费 Deep Learning Academy 课时。 这是第 2 节课，共 4 节。 你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。 这是 Deep Learning Academy 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 Deep Learning Academy 课程共包含 4 节课。
本课时的部分内容尚未翻译，以英文显示。
One Filter Size to Rule
VGG made a bold bet: use only tiny 3x3 convolutions everywhere, and just stack a lot of them.
Why Small Wins
Two stacked 3x3 layers see the same area as one 5x5, but with fewer parameters and an extra nonlinearity in between.
Depth Over Width
Instead of fat layers, VGG chose to go deep, reaching 16 or 19 layers and learning richer feature hierarchies.
Uniform and Predictable
Every conv keeps the same padding and size, so shapes stay easy to reason about. This uniformity made VGG simple to copy.
Pool to Shrink
After each block of convs, a 2x2 max pool halves the spatial size while doubling the channel count.
import torch.nn as nn
pool = nn.MaxPool2d(kernel_size=2, stride=2)A VGG Block
The repeating unit is conv, ReLU, conv, ReLU, pool. Chain these blocks and you have most of VGG.
import torch.nn as nn
block = nn.Sequential(nn.Conv2d(64, 128, 3, padding=1), nn.ReLU(), nn.MaxPool2d(2))A Heavy Classifier Head
VGG ends with three large dense layers. They hold most of its weights, which is why the model is so big on disk.
The Memory Cost
That elegance has a price: VGG is slow and memory-hungry. Beauty in design did not mean cheap to run.
A Strong Feature Extractor
Even today, VGGs early layers make excellent reusable features for transfer learning and style transfer.
Build It with Loops
Because the pattern repeats, you can generate VGG from a config list, looping to add each layer programmatically.
cfg = [64, 64, "M", 128, 128, "M"]
# "M" means insert a max-pool hereThe Lasting Lesson
VGGs gift was a clear principle: prefer many small filters over a few large ones. Modern nets still follow it.
Quick Check
Recall the central design choice behind VGG.
Recap: Small but Mighty
VGG proved that depth built from simple 3x3 blocks beats clever big filters. Simple, repeatable, and influential. Well done!
常见问题解答
「VGG：小型滤波器的堆叠」课时是免费的吗？
是的 — 「VGG：小型滤波器的堆叠」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 Deep Learning Academy 课程的其余内容，请升级到 CoddyKit PRO。 Deep Learning Academy 课程共包含 4 节课。
「VGG：小型滤波器的堆叠」这节课中我会学到什么？
通过简单的 3x3 模块构建深度 你通过在浏览器中直接运行的动手代码来练习 Deep Learning Academy，全天候 AI 导师会在你学习这节课的过程中回答你的问题。
学习 Deep Learning Academy 需要有经验吗？
无需任何先前经验。CoddyKit 上的 Deep Learning Academy 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。 这是第 2 节课，共 4 节。
「VGG：小型滤波器的堆叠」课时需要多长时间？
大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。
我能在这节 Deep Learning Academy 课中编写并运行代码吗？
能。每节 Deep Learning Academy 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。