| CNN | Convolutional neural network |
| HRSuperNet | High-resolution super network |
| S-HRNet | Slimmed network from HRSuperNet |
| SD-HRNet | Refined S-HRNet using triple knowledge distillation |
| LSFF | Lightweight selective feature fusion |
| TA | Transformation and aggregation |
| MBConvs | Mobile inverted bottleneck convolutions |
| KD | Knowledge distillation |
| KL | Kullback–Leibler |
| 2D | Two-dimensional |
| FLOPs | Number of floating-point operations |
| NME | Normalized mean error |
| #Params | Number of parameters |
| FPS | Frames per second |
| SIFT | Scale-invariant feature transform |
| M | Mega |
| G | Giga |