# -10\1進度: ###### tags:`Meeting` ## 研究與學習: ### Academic Approach: * Survey paper: * [Cross-modalitysynthesis from CT to PET using FCN and GAN networks for improved automated lesion detection](https://hackmd.io/twu3TipgSamODHlEmdqZXA) * [Denoising Diffusion Probabilistic Models](https://arxiv.org/pdf/2006.11239.pdf) * Markov chain-->當前的任何狀態轉變的機率,都與過去狀態完全無關 *  * 學習逐步denoise * 在訓練模型時,逐步地將原圖加入雜訊。具體上來說,使用一個高斯分布一次又一次地在原圖上打上很小的雜訊,然後讓網路來學習如何reverse這個雜訊。(?) *  * https://openai.com/dall-e-2/ * https://imagen.research.google/ ## 其他: * 鼓勵藍放射師: ) * 多益習題第一回 * [帳務](https://hackmd.io/xhAzrk2zTX2hbp5zXLuepQ) * 為什麼Normal Distribution這麼好用? * Q:為什麼Normal Distribution通常作為雜訊的分布? * Q:為何在DL(deep learning),參數的初始化要用Normal Distribution? * Q:為何在Bayesian公式裡常常會使用Normal Distribution當作Prior Probability? * Q:在使用GAN(generative adversarial network)時,為什麼給予的輸入要假設Normal Distribution? # 下週進度: ###### tags:`Meeting` ## 研究與學習: ### Academic Approach: * Survey paper: * [Denoising Diffusion Probabilistic Models](https://arxiv.org/pdf/2006.11239.pdf) * [Attention UNet](https://arxiv.org/abs/1804.03999) * [Generative Multiplane Images: Making a 2D GAN 3D-Aware](https://arxiv.org/pdf/2207.10642.pdf)ECCV * [Attention U-Net:Learning Where to Look for the Pancreas](https://hackmd.io/twu3TipgSamODHlEmdqZXA) * [Point Cloud GAN](https://arxiv.org/pdf/1810.05795.pdf) * [Mult-view]() ## 其他: * 多益習題第2回 * [帳務](https://hackmd.io/xhAzrk2zTX2hbp5zXLuepQ) * 等藍放射師訓練3DCNN模型
×
Sign in
Email
Password
Forgot password
or
By clicking below, you agree to our
terms of service
.
Sign in via Facebook
Sign in via Twitter
Sign in via GitHub
Sign in via Dropbox
Sign in with Wallet
Wallet (
)
Connect another wallet
New to HackMD?
Sign up