2199
Future Blog Post
Published:
This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.
2026
Open-sourcing my first game: AI grids improved the workflow, not the file size
Published:
I open-sourced Absorbed, the first game I have built, as a clean Apache 2.0 repository. It is an ongoing landscape iPhone and iPad roguelike inspired by Darkest Dungeon, Slay the Spire, and creature-collecting games, reinterpreted through Buddhist ideas about attachment, identity, and letting go.
Building Daniel OS: data, training, and strict evaluation
Published:
Daniel OS is a personal portfolio assistant that runs its generative model in the visitor’s browser. It combines a verified profile index, a cited entity index, optional public retrieval, a personalized language model for conversational synthesis, browser speech recognition, and local speech output. The goal is to make the portfolio queryable without allowing personalization to distort ordinary technical knowledge.
2020
OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
Published:
Introduction
Review of MobileNet
Published:
Study of MobileNet and basic information and coding of Cov2d, Forward propagation, BN, Quantized, etc…
Lightweight 3D Human Pose Estimation Network Training Using Teacher-Student Learning
Published:
Review of paper named ‘Lightweight 3D Human Pose Estimation Network Training Using Teacher-Student Learning’
YOLACT++
Published:
Review of paper named YOLACT++: Better Real-time Instance Segmentation
PoseFlow: Efficient Online Pose Tracking
Published:
Basically estimation and tracking is easy terminology by understanding definition of respective keywords. Additionally, in a viewpoint of each results, both may same exactly although have different meaning. There are two methods of tracking human pose, one is top-down, the other is bottom-up. In my comprehension estimation is usually used in an image and tracking is used in a video.