| YOLACT-550 (ResNet-101-FPN) | 46.6 | 29.2 | 44.8 | 29.3 | 9.2 | 33.3 (Titan Xp) | 28.2 | YOLACT: Real-time Instance Segmentation | |
| YOLACT-550++ (ResNet-101-FPN) | 53.8 | 36.9 | 55.1 | 36.8 | 11.9 | 27.3 (Titan Xp) | 34.6 | YOLACT++: Better Real-time Instance Segmentation | |
| SparseInst-608 (ResNet-50-vd) | 59.2 | 40.2 | 56.9 | 39.4 | 15.7 | 40 (2080 Ti) | 37.9 | Sparse Instance Activation for Real-Time Instance Segmentation | - |
| maYOLACT-700 (ResNet-50) | 59.4 | 39.9 | 52.5 | 40.8 | 18.1 | 25 (Tesla V100) | 37.7 | Mask-aware IoU for Anchor Assignment in Real-time Instance Segmentation | |
| SipMask++ (ResNet-101, single-scale test) | 55.6 | 37.6 | 56.8 | 38.3 | 11.2 | 27.0 (Titan Xp) | 35.4 | SipMask: Spatial Information Preservation for Fast Image and Video Instance Segmentation | |
| SipMask (ResNet-50, single-scale test) | 51.9 | 32.3 | 49.8 | 33.6 | 9.2 | 41.7 (Titan Xp) | 31.2 | SipMask: Spatial Information Preservation for Fast Image and Video Instance Segmentation | |
| maYOLACT-550 (ResNet-50) | 56.2 | 37.1 | 51.4 | 38.0 | 14.7 | 30 (Tesla V100) | 35.2 | Mask-aware IoU for Anchor Assignment in Real-time Instance Segmentation | |
| SipMask (ResNet-101, single-scale test) | 53.4 | 34.3 | 54.0 | 35.6 | 9.3 | 31.3 (Titan Xp) | 32.8 | SipMask: Spatial Information Preservation for Fast Image and Video Instance Segmentation | |
| BlendMask-512 (DLA_34) | - | - | - | - | - | 33.3 | 35.2 | BlendMask: Top-Down Meets Bottom-Up for Instance Segmentation | |
| SOLO-512 | 57.7 | 39.7 | - | - | - | 31.3 | 37.1 | SOLOv2: Dynamic and Fast Instance Segmentation | |
| CenterMask-Lite (ResNet-50-FPN) | - | - | 48.7 | 34.7 | 12.9 | - | 32.9 | CenterMask : Real-Time Anchor-Free Instance Segmentation | |
| SparseInst-448 (ResNet-50-vd) | 56.5 | 37.7 | 57.0 | 37.1 | 12.3 | 58.5 (2080 Ti) | 35.9 | Sparse Instance Activation for Real-Time Instance Segmentation | - |
| Predicting Future Instance Segmentation | - | - | - | - | - | 8 | 40.2 | - | - |