Table 2.
Classification results using CNN for different scenarios.
| Training | Testing | Recognition | Accuracy |
|---|---|---|---|
| One person walking (128) | One person walking (80) | Person | 78.75–86.25% |
| One person walking on different tracks (32) | One person walking on different tracks (20) | Track | 85.00–100.00% |
| One person walking with different shoes (32) | One person walking with different shoes (20) | Shoes | 80.00–100.00% |
| One person walking and pushing a cart (32) | One person walking and pushing a cart (20) | Cart | 93.75–100.00% |
| One person running (128) | One person running (80) | Person | 76.25–86.25% |
| One person running on different tracks (32) | One person running on different tracks (20) | Track | 95.00–100.00% |
| One person running with different shoes (32) | One person running with different shoes (20) | Shoes | 75.00–90.00% |
| One person walking (208) | Two people walking (104) | Person |
63.85–78.46% (either) 28.46–44.62% (both) |
| One person running (208) | Two people running (78) | Person |
60.26–76.92% (either) 34.62–50.00% (both) |
| One person walking + one person running (416) | One person walking and another running (104) | Person and movement |
54.62–69.23% (either) 46.92–61.54% (both) |
The number of datasets for training and testing, respectively, is included in parentheses.