| 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 |
| 31 |
Tags
- object detection
- facenet
- ccdf
- Ai
- f2dnet
- Triplet
- deepfake detection
- linear inverse problem
- ssda-yolo
- ddpm
- hqs-algorithm
- ilvr
- fourmer
- neighboring pixel relationships
- unconditional generative models
- semi-supervied learning
- diffusion
- ddrm
- ML
- daod
- unconditional ddpm
- image restoration
- conditional diffusion
- stochastic contraction theory
- focal detection network
- diffpir
- IR
- pedestrian detection
- face forgery detection
- ddim
Archives
- Today
- Total
목록f2dnet (1)
Stand on the shoulders of giants
Main ideaFocal detection network + fast/light-weight suppression head(based on CSP)Focal detection network ⇒ rpn w/ per pixel center & scale regressionlight suppression head ⇒ detached settings ∴ lightFalse Positive를 줄인다. (by fast suppression)*기존 : RPN + Detection head(bbox heads)Anchor free method*기존: single/multi-stage detectors는 rely on anchorsloss MR-2 in havy occlusion settings BackgroundP..
Paper reviews
2024. 5. 14. 14:25