Research Article
Motion Recognition Based on Deep Learning and Human Joint Points
Table 2
Anchor candidate box size based on the K-means clustering algorithm.
| Candidate box number | Candidate box size (width ∗ height unit: pixel) |
| Candidate box 1 | 18 × 44 | Candidate box 2 | 22 × 55 | Candidate box 3 | 31 × 76 | Candidate box 4 | 43 × 105 | Candidate box 5 | 59 × 145 | Candidate box 6 | 82 × 200 | Candidate box 7 | 113 × 276 | Candidate box 8 | 156 × 381 | Candidate box 9 | 215 × 526 | Candidate box 10 | 45 × 20 | Candidate box 11 | 110 × 42 | Candidate box 12 | 250 × 100 | Candidate box 13 | 500 × 204 |
|
|