Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Sign up for free
Menu
Search
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Pricing
Search
Sign in
Sign up for free
20190904 progress report
Search
Scott Li
September 04, 2019
70
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
20190904 progress report
Scott Li
September 04, 2019
More Decks by Scott Li
See All by Scott Li
20190516 progress report
107368024
0
47
20190411 progress report
107368024
0
50
20190221 progress report
107368024
0
37
20181213 progress report.pdf
107368024
0
74
20181101 progress report
107368024
0
110
Featured
See All Featured
Reality Check: Gamification 10 Years Later
codingconduct
0
2.3k
<Decoding/> the Language of Devs - We Love SEO 2024
nikkihalliwell
1
360
JavaScript: Past, Present, and Future - NDC Porto 2020
reverentgeek
52
6.1k
The Illustrated Guide to Node.js - THAT Conference 2024
reverentgeek
1
570
The AI Search Optimization Roadmap by Aleyda Solis
aleyda
2
6.3k
Principles of Awesome APIs and How to Build Them.
keavy
128
18k
DevOps and Value Stream Thinking: Enabling flow, efficiency and business value
helenjbeal
2
410
Fireside Chat
paigeccino
43
4.1k
Taking LLMs out of the black box: A practical guide to human-in-the-loop distillation
inesmontani
PRO
3
2.5k
The Impact of AI in SEO - AI Overviews June 2024 Edition
aleyda
6
1.2k
SEO in 2025: How to Prepare for the Future of Search
ipullrank
3
3.9k
Building Experiences: Design Systems, User Experience, and Full Site Editing
marktimemedia
1
630
Transcript
108/09/04 進度報告 學⽣生:李政璋 教授:曾柏軒
108/09/04 進度報告 【⽬目前進度 】環境配置 2 ‣ 資料數:200 對⼿手動標註影像 ‣ Model:
U-Net ‣ Input shape: 512 * 512 * 1 ‣ Output shape: 512 * 512 * 1 ‣ Batch size: 2 ‣ Epoch: 50 ‣ GPU: NVIDIA GeForce GTX 1080 Ti 11GB
108/09/04 進度報告 【⽬目前進度 】訓練資料 ▸ Training set example 3 Original
image Label image
108/09/04 進度報告 【⽬目前進度 】初始訓練結果 4 ▸ Loss function 為 MSE,並改變
label data,得到增強的⾎血管影像。
108/09/04 進度報告 【⽬目前進度 】分割訓練結果 5 ▸ Loss function 為 binary
cross entropy,並改變 label data 為⽩白底⿊黑值, 得到 segmentation 的影像。 sen: 靈敏度,⾎血管的完整性。⾎血管越完整,靈敏度越⾼高。 spe: 特異異度,雜訊的程度。雜訊越低,特異異度越⾼高。 pre: 精確度,訓練出來來的影像,判斷正確的比例例。
108/09/04 進度報告 【⽬目前進度 】不同的訓練集 ▸ 訓練出來來的模型,套⽤用在醫師助理理標註的影像。 6
108/09/04 進度報告 【⽬目前進度 】比較原始⽅方法 ▸ 與傳統的 frangi filter 比較,靈敏度較⾼高,特異異度則降低。 7
108/09/04 進度報告 問題描述 1. Deeplab Tutorial 套⽤用醫學影像資料集,無法跑出訓練結果。 8
108/09/04 進度報告 本⽉月預計進度 1. 提⾼高訓練資料數量量。 2. 解決 Deeplab 無結果的問題。 9
THANK YOU FOR YOUR LISTENING!