joshvarty / cancerdetection Goto Github PK
View Code? Open in Web Editor NEWWorking on: https://www.kaggle.com/c/histopathologic-cancer-detection
License: MIT License
Working on: https://www.kaggle.com/c/histopathologic-cancer-detection
License: MIT License
We should use:
We could try using weight decay to reduce overfitting.
We only care about tumors in the 32x32 center of the image. Maybe we should try cropping to 48x48.
It might help to have a custom head with an extra fully connected layer.
Apparently I have been using the wrong metric. Let's fix that.
We need to create a proper validation set by removing sets of images that came from the same original slide picture. After creating the validation set we should train a few networks on it and compare the validation AUC score to what we get on the leaderboard.
When we increase validation score, does our LB usually improve?
It's possible normalization made our performance worse.
slice(1e-6,1e-2))
(Cancer Normalization)Name | Valid | Test |
---|---|---|
Resnet34 | 92.7 | 93.4 |
Resnet50 | 95.4 | 94.1 |
Resnet101 | 96.1 | 94.1 |
Resnet152 | 94.5 | 0.918 |
SqueezeNet1_1 | 83.3 | 0.8970 |
Densenet121 | 95.1 | 92.9 |
slice(1e-5,2e-2)
(ImageNet normalization)Name | Valid | Test |
---|---|---|
Resnet50 | 96.9 | 95.8 |
slice(1e-6, 1e-2)
(Imagenet normalization)Name | Valid | Test |
---|---|---|
Resnet50 | 97.6 | 94.29 |
Resnet101 | 97.7 | 94.87 |
Resnet152 | 97.6 | 95.23 |
Densenet121 | 97.6 | 95.19 |
slice(1e-6, 1e-2)
(Updated cancer normalization)Name | Valid | Test |
---|---|---|
Resnet50 | 97.5 | 94.57 |
Resnet101 | 97.8 | 94.66 |
Resnet152 | 97.6 | 94.75 |
Densenet121 | 97.4 | 95.01 |
slice(1e-6, 1e-2)
+ 5 epochs @ (1e-7, 1e-3) (Updated cancer normalization)Name | Valid | Test |
---|---|---|
Resnet50 | 97.4 | 94.93 |
Resnet101 | 97.5 | 94.44 |
Resnet152 | 97.8 | 94.55 |
Densenet121 | 97.6 | 95.13 |
slice(1e-6, 1e-2)
(Imagnet normalization) More transformationsName | Valid | Test |
---|---|---|
Resnet50 | 97.84 | - |
Resnet101 | 97.76 | 95.42 |
Resnet152 | 98.21 | 95.70 |
Densenet121 | ||
Densenet161 |
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.