# Pelee: Real-time Object Detection System on Mobile Devices

The rise of deep learning in the past decade has been astronomical, especially after introduction of CNN(Convolutional Neural Network). But this rise has been accompanied with bigger models and need for large compute power. Often these large(and compute heavy) models are hard to deploy for real-life application, especially on edge-devices…

# Segmentation in Robotic Surgery

Application of Deep Learning in Robotic Surgery

This is my second blog based on the fastai lesson. I wanted to apply the learning from the lesson to a different dataset with binary labels. …

# Are you Chinese, Japanese or Korean?

Image Classifier based on fast.ai lecture

One of the challenges that I face living in S. Korea is to tell the difference between Chinese, Japanese and Koreans. The similarities in their appearances has led to many awkward moments during my stay.

I wish to avoid those awkward moments by building…

# Solving Markov Decision Process

Policy Iteration+ Value Iteration

In the last post, I wrote about Markov Decision Process(MDP); this time I will summarize my understanding of how to solve MDP by policy iteration and value iteration.

So what is policy iteration and value iteration?

These are the algorithms in Dynamic Programming that are used…

# Gradient Descent: Stochastic vs Batch

The difference is in weight updating patterns

This is my first post on #100DaysofMLCode challenge; everyday I plan to read(which will result in blog to check my understanding) and/or coding.

In this post, I will talk about Stochastic Gradient Descent and its difference with Batch Gradient Descent. I will also…

# Markov Decision Process(MDP) Simplified

MDP gives the mathematical formulation of Reinforcement Learning Problem

Markov Decision Process(MDP) is an environment with Markov states; Markov states satisfy the Markov Property: the state contains all the relevant information from the past to predict future. Mathematically,

So if I say that the state S<t> is Markov, that means…

# Spine Segmentation using U-Net

This post is based on my internship experience where I worked on the segmentation of Spine using U-Net architecture.

Data-Set: CT scans of 11 patients collected from the institution-affiliated hospital. The data were in dicom format with no labels.