-
作者您好,首先很感谢您优秀的工作,在这里有一个问题想问您,请问训练过程中咱判别器损失的变化曲线是什么样的,我将本判别器模型迁移到一个VAE重建的应用场景,在训练过程中,无论是否加载您的判别器pretrained权重,GAN的损失均基本保持不变,请问这是正常的吗?期待您的回复,对您感激不尽!
-
Hi,
I am little bit confused after reading your paper. Please correct me if I misunderstood.
In your paper, you show difference between VAE and WAE in terms of distribution matching objective
* V…
j-min updated
6 years ago
-
Hi!
It is a great pleasure to meet an incredible work!
I appreciate your dedication to this community.
I have checked the paper and github, but it seems that there is no sample generation perform…
-
## 一言でいうと
ブラシストロークを学習させるネットワークの提案。VAEで画像の表現を学習、全結合でブラシアクションをVAE表現との対応を学習、GANでブラシアクションによる画像生成を学習、と3段階で学習を行い学習効率をあげている。生成画像のクラス分類タスクを入れることで、任意クラスの画像生成も行なっている
### 論文リンク
https://arxiv.org/abs/190…
-
Dear sir:
I don't know why you add "the feature matching loss".In your paper, you say that "We compare our final implementation to the results using (1) only GAN loss, and (2) GAN + feature m…
-
- [ ] モデルの選定
- [x] 学習データの作成
- [ ] プログラム作成
-
## 一言でいうと
VAEやGANなどの生成系のタスクでは、「真の分布」との距離の最小化を目的にしている。つまり「距離」の定義はモデルの精度の大きな要素で、GANではこの自由度が高い反面学習が安定しない要因になっていた。そこでWasserstein距離を使うと勾配が消失せず学習が安定したという話。
### 論文リンク
https://arxiv.org/abs/1701.0787…
-
VAE turns to be a quite good model when we want to generate similar yet new images. However, as our data has several patterns that is hard to capture, such as the geographical features, the delicate c…
-
Do you have any saved checkpoints of your model trained on CelebA (or any other face dataset) that you could share? Thanks!
-
The following section contains the math of latent code generation.
Where can we find " what exact math operation does it perform " ?
Thanks
====
def reparameterize(self, mu, logvar):
…