构建 DCGAN
用于图像生成的卷积 GAN
构建 DCGAN 是 CoddyKit 上的免费 Deep Learning Academy 课时。 这是第 3 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。
What DCGAN Adds
A DCGAN is a GAN built from convolutional layers. It swaps dense layers for convolutions, which makes it great at generating realistic images. 🖼️
Generator Grows the Image
The generator uses transposed convolutions to upsample a tiny latent vector step by step into a full-size image.
nn.ConvTranspose2d(100, 256, 4, 1, 0)Discriminator Shrinks the Image
The discriminator does the reverse: strided convolutions shrink the image down to a single real-or-fake score.
nn.Conv2d(3, 64, 4, 2, 1)Batch Norm Stabilizes
DCGAN puts batch normalization between layers in both networks. It keeps activations healthy and makes training far more stable.
nn.BatchNorm2d(256)ReLU in the Generator
The generator uses ReLU activations on its hidden layers to keep gradients flowing as the image is built up.
nn.ReLU(True)LeakyReLU in the Discriminator
The discriminator prefers LeakyReLU, which lets a small signal through for negative values and avoids dead neurons.
nn.LeakyReLU(0.2, inplace=True)Tanh on the Output
The generator ends with Tanh, squeezing pixels into the range minus one to one to match normalized training images.
nn.Tanh()Sigmoid for the Verdict
The discriminator finishes with a sigmoid, turning its final feature into a probability between zero and one.
nn.Sigmoid()Weight Initialization
DCGAN starts weights from a small normal distribution centered near zero. Good init helps both networks train smoothly from the first step.
nn.init.normal_(m.weight, 0.0, 0.02)Pick the Latent Size
The latent dimension, often 100, sets how many random numbers seed each image. Larger latents give the generator more room to vary.
nz = 100Watch Samples as You Train
Save a grid of generated images every few epochs. Watching them get sharper is how you confirm a DCGAN is actually learning.
Quick Check
Which activation belongs at the very end of the generator?
Recap: Building a DCGAN
You assembled a DCGAN: transposed convolutions grow images, strided convolutions judge them, and batch norm keeps it all stable. 🛠️
用 AI 导师学习 Python — 免费
在浏览器中编写并运行真实代码，获得全天候 AI 导师的即时帮助，并在网页或应用中继续学习。
- 课程
- 30
- 课程
- 120
常见问题解答
「构建 DCGAN」课时是免费的吗？
是的 — 「构建 DCGAN」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。
「构建 DCGAN」这节课中我会学到什么？
用于图像生成的卷积 GAN 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 3 节课，共 4 节。
「构建 DCGAN」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。
此课程中的所有课时
- 生成器与判别器：这场博弈
- 对抗损失
- 构建 DCGAN
- 模式崩溃与稳定化技巧