Table 3.
Comparison of the proposed ER method with state-of-the-art methods for 19 + 1 classes scenario. Best results are written in bold number.
| Methods | Aero | Bike | Bird | Boat | Bottle | Bus | Car | Cat | Chair | Cow | Din. Table | Dog | Horse | M Bike | Person | Plant | Sheep | Sofa | Train | TV | mAP at 0.5 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| All data | 76.5 | 84.4 | 68.9 | 59.3 | 55.6 | 82.1 | 85.9 | 79.9 | 55.4 | 73.7 | 67.8 | 79.9 | 85.2 | 84.4 | 83.1 | 46.1 | 71 | 70.3 | 82.1 | 71.7 | 73.2 |
| KD [1] | 50.2 | 53.0 | 29.8 | 27.5 | 13.3 | 55.4 | 67.2 | 45.5 | 13.3 | 36.9 | 34.2 | 38.7 | 60.7 | 56.3 | 38.5 | 10.5 | 33.8 | 33.0 | 55.6 | 15.8 | 38.4 |
| DMC [17] | 77.2 | 67.8 | 63.5 | 46.5 | 50.2 | 70 | 79.3 | 80.3 | 43.5 | 65.5 | 0.2 | 71.1 | 74.6 | 71.5 | 78.2 | 39.6 | 66.6 | 34.8 | 79.2 | 42.4 | 60.1 |
| GEM [15] | 66.6 | 72.6 | 54.9 | 48.1 | 46.7 | 72.5 | 80.5 | 68.7 | 37.9 | 66.3 | 48.4 | 62.5 | 81 | 72.5 | 67.3 | 33.5 | 68 | 45.8 | 76.9 | 28.4 | 60 |
| ER* | 73.2 | 83 | 62.9 | 56.3 | 53.7 | 81.1 | 84.9 | 77.7 | 48.4 | 69.1 | 63.6 | 78.5 | 85.7 | 82.6 | 80.9 | 43.3 | 67.7 | 60.8 | 79.6 | 39.7 | 68.6 |
| ER** | 76.7 | 83.1 | 62.5 | 56.8 | 53.2 | 83.8 | 85 | 78.1 | 50.1 | 68.6 | 62.1 | 77.4 | 85.6 | 79.7 | 80.7 | 42.8 | 70.6 | 66.2 | 81.1 | 33.3 | 68.9 |
| ER*** | 74.3 | 79.5 | 59.5 | 55.3 | 49.5 | 78.9 | 83.5 | 77.1 | 45.2 | 66 | 57.6 | 73.9 | 85.5 | 80.3 | 78.3 | 40.1 | 68 | 58.1 | 78.7 | 44.7 | 66.7 |
ER* method without augmentation before uploading to memory (2500), ER** method with augmentation after uploading to memory (2500), ER*** method with augmentation after uploading to memory (1000).