This repository is dedicated to reviewing academic papers in machine learning, deep learning, and computer vision.
The goal is not just to summarize, but to critically analyze, reconstruct key ideas, and explore potential applications or implementations.
It was created as part of my effort to read academic papers regularly, and to systematically document my understanding, thoughts, and experiments based on them.
Many folders include not just summaries, but also comparative experiments or personal implementations inspired by the paper's content.
- Build a structured habit of reading and understanding academic papers
- Translate complex ideas into my own words
- Re-implement key methods or concepts when possible
- Strengthen research mindset and prepare for graduate school
- Test and adapt paper methods to my own datasets and goals
Each review typically includes:
- Background and motivation
- Problem definition and main contributions
- Explanation of key models or algorithms
- Analysis of experiments and results
- Limitations and possible future directions
- Optional: Personal implementations or variations
- Machine Learning (e.g., ensemble methods, regularization, optimization)
- Deep Learning (e.g., CNNs, transformers, attention mechanisms)
- Computer Vision (e.g., pose estimation, action recognition)
- Applied AI (e.g., healthcare, recommendation systems)
- 01_dowhy
DoWhy: An End-to-End Library for Causal Inference (NeurIPS 2020 Workshop)
Includes: Summary (multi-day notes), implementation notebook, refutation experiments
-
01_mobilenetv2
MobileNetV2: Inverted Residuals and Linear Bottlenecks (CVPR 2018) -
02_efficientnet
EfficientNet: Rethinking Model Scaling for CNNs (ICML 2019) -
03_dance_action_recognition
Action Recognition using Pose Estimation (2019) -
04_explainable_martial_arts_eval
Explainable Quality Assessment of Skeletal Representations for Martial Arts Movements (2024) -
05_few_shot_grounding_dino_agri
Few-Shot Grounding DINO for Agriculture (2023)
- 01_attention_is_all_you_need
Attention Is All You Need (NeurIPS 2017)
Includes: Day-by-day summaries (abstract, method, experiments, conclusion), math notes, and implementation details
- π Aim to review at least one paper per week
- π Include code snippets or full re-implementations when possible
- π Organize reviews by topic (e.g., CV, NLP, theory)
- π Conduct small-scale experiments inspired by each paper
- GitHub: @hojjang98
- Blog: hojjang98.github.io
π§ This repository is a work in progress.
I welcome discussions, suggestions, and collaboration from fellow learners and researchers!