| 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
- ML
- conditional diffusion
- ssda-yolo
- facenet
- ddpm
- deepfake detection
- unconditional ddpm
- fourmer
- neighboring pixel relationships
- object detection
- pedestrian detection
- Ai
- face forgery detection
- ddrm
- semi-supervied learning
- focal detection network
- diffpir
- IR
- f2dnet
- daod
- ilvr
- linear inverse problem
- hqs-algorithm
- image restoration
- Triplet
- ddim
- ccdf
- diffusion
- unconditional generative models
- stochastic contraction theory
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