Notice
Recent Posts
Recent Comments
Link
일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | ||
6 | 7 | 8 | 9 | 10 | 11 | 12 |
13 | 14 | 15 | 16 | 17 | 18 | 19 |
20 | 21 | 22 | 23 | 24 | 25 | 26 |
27 | 28 | 29 | 30 |
Tags
- DP
- CNN
- BFS
- C++
- object detection
- eecs 498
- Reinforcement Learning
- LSTM
- 딥러닝
- 머신러닝
- real-time object detection
- Mask Processing
- dfs
- opencv
- machine learning
- Python
- YoLO
- two-stage detector
- 그래프 이론
- One-Stage Detector
- MinHeap
- canny edge detection
- r-cnn
- deep learning
- MySQL
- dynamic programming
- 강화학습
- image processing
- 백준
- AlexNet
Archives
- Today
- Total
목록universal approximation ability (1)
JINWOOJUNG

PaperPointNetIntroduction 기존 Deep Learning 연구에서의 CNN은 Regular Input Data를 기반으로 동작한다(Image, Voxel, etc). 하지만, Point Cloud는 순서가 없는(Inordered) Inrregular Data이다. 따라서 Raw Point Cloud Data(PCD)를 CNN의 Input으로 취하는 것은 한계가 있다. 전처리 과정에서 PCD를 Regular Form(3D Voxel Grid, Collections of Images, etc)으로 변환하는 과정은 불필요한 메모리 및 계산 시간 증가를 초래하며, Quantization Artifacts에 의한 원본 정보 왜곡 및 Natural Invariances이 깨질 수 있다. Qua..
딥러닝/논문
2025. 2. 3. 23:20