- "μμ¨μ£Όν κΈ°μ κ°λ° νμ μ¬μ
"μ ν΅ν΄ ꡬμΆν μμ¨μ£Όν κ³΅κ° λ°μ΄ν°μ
μ νμ©νμ¬ μμ¨μ£Όνμ°¨-μΈνλΌ μ°κ³ν AI κΈ°μ κ°λ°
- μ£Όννκ²½μμ μΉ΄λ©λΌ μΌμλ₯Ό μ΄μ©νμ¬ μ νΈλ± μΈμ
:sunglasses:νμ μκ°
λ°μ΄ν°μ
μ 보
λ°μ΄ν°μ
ν΅κ³
- νμ΅ λ°μ΄ν°: 26,864 νλ μ
- νκ° λ°μ΄ν°: 13,505 νλ μ
λ°μ΄ν° νν
| μ νΈλ± μ ν |
ν΄λμ€ μ΄λ¦(λ²νΈ) |
μ νΈλ± μ ν |
ν΄λμ€ μ΄λ¦(λ²νΈ) |
| μ°¨λ μ νΈλ± |
Go (0) |
보νμ μ νΈλ± |
Go (7) |
|
GoLeft (1) |
|
NoSign (8) |
|
NoSign (2) |
|
Stop (9) |
|
Stop (3) |
λ²μ€ μ νΈλ± |
Go (10) |
|
StopLeft (4) |
|
NoSign (11) |
|
StopWarning (5) |
|
Stop (12) |
|
Warning (6) |
|
Warning (13) |
Models
| Model |
Backbone |
Pre-trained |
Epochs |
oversampling |
Image size |
val mAP50 |
| yolo |
C3K2 |
yolo11x |
20 |
X |
1280x1280 |
0.6010 |
| Co-DINO |
Swin-L |
Object365,COCO |
1 |
X |
1024x1024 |
0.6407 |
| Co-DINO |
Swin-L |
Object365,COCO |
2 |
X |
1024x1024 |
0.6821 |
| Co-DINO |
Swin-L |
Object365,COCO |
3 |
X |
1024x1024 |
0.6833 |
| Co-DINO |
Swin-L |
Object365,COCO |
1 |
O |
1024x1024 |
0.6990 |
| Cascade-RCNN |
Swin-L |
COCO |
5 |
X |
1024x1024 |
0.6819 |
| Cascade-RCNN |
Swin-L |
COCO |
2 |
X |
1024x1024 |
0.6390 |
| Cascade-RCNN |
Swin-L |
COCO |
2 |
O |
1024x1024 |
0.6875 |
This project is released OpenMMLab / λͺ¨λΈ μ±λ₯ν
Ensemble
| μμλΈ κΈ°λ² |
Co-DINO 1ep over |
Co-DINO 1ep |
Co-DINO 2ep |
Co-DINO 3ep |
Cascade-RCNN 5ep |
Cascade-RCNN 2ep |
Cascade-RCNN 2ep over |
Test mAP50 |
| NMW |
|
o |
o |
o |
|
o |
|
0.6945 |
| NMW |
o |
|
|
o |
o |
|
o |
0.7344 |
| Classwise |
o |
o |
o |
o |
o |
|
o |
0.7362 |
Results
How to use
