craftsangjae / deeplearning-dojo

good code is simple. Implement Various DeepLearning Algorithms such as VAE, MC Dropout, Grad-CAM, Object Detection, Semantic Segmentation and so on

Objective

Code Kata is defined as an exercise in programming which helps hone our skill through practice and repetition. In machine learning programming, Code Kata for implementing ML algorithms is very important, becuase we can realize the details ( such as Data Sampling, Weight initialization, various training strategy ...) while implementing the algorithm.

I implement various algorithms using in Deep Learning and organize them into scripts.. I'll update one script each week.

If you have a good topic, feel free to leave it on the issue! I will try to implement it as much as possible!

How to do the Code Kada together? (set-up environment)

Do not worry! I provide the environment written as a docker image.

# Run it From the root project directory
docker-compose up -d

Deep-Learning Katas List

---

Embedding Network for Handling Categorical Variables

Goals

• Implement a simple Deep Learning Model Handling variables by Embedding Layer
• Acheive Accuracy similar to that of a Random Forest Classifier

Dataset

• ucl - bank marketing data set

---

Maxout Network using Tensorflow

Goals

• Implement Maxout Activation Layer using Keras custom Layer
• Check the performance by learning fashion-MNIST

Dataset

• Fashion-MNIST

papers

• Maxout Networks

---

two ways for visualizing CNN, guided back propagation and GradCAM

Goals

• Implement two methods to interpret the output of CNN

• Algorithm implementation based on the operation of Resnet50v2

papers

• Striving for Simplicity: The All Convolutional Net

• Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization

---

CRNN(Text Recognition) Using Tensorflow <--TODO

Goals

• Implement CRNN, Deep Learing Model for reading text in images.

Datasets

• synthetic dataset using MNIST

Papers

• An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition

---

Shake-Shake Regularization using Tensorflow <-- TODO

Goals

• Implement Shake-Shake Regularization which directly applies noise inside the Deep Learning Model

Datasets

• CIFAR-10

Papers

• Shake-Shake Regularization

CopyRight

This repository is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.