Visual Reasoning on CLEVR 1.0 (test)
97.7Overall AccuracyCNN+GRU+FiLM
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CNN+GRU+FiLMtraining from raw pixels=false2017.09 | 97.7 | 94.3 | 99.1 | 96.8 | 99.1 | 99.1 | |
| CNN+GRU+FiLMtraining from raw pixels=true2017.09 | 97.6 | 94.3 | 99.3 | 93.4 | 99.3 | 99.3 | |
| PG+EEextra supervision via program labels=true, program labels=700K2017.09 | 96.9 | 92.7 | 97.1 | 98.7 | 98.1 | 98.9 | |
| CNN+LSTM+RNdata augmentation=true, training from raw pixels=true2017.09 | 95.5 | 90.1 | 97.8 | 93.6 | 97.9 | 97.1 | |
| Human2017.09 | 92.6 | 86.7 | 96.6 | 86.5 | 95 | 96 | |
| PG+EEextra supervision via program labels=true, program labels=9K2017.09 | 88.6 | 79.7 | 89.7 | 79.1 | 92.6 | 96 | |
| N2NMNextra supervision via program labels=true2017.09 | 83.7 | 68.5 | 85.7 | 84.9 | 90 | 88.7 | |
| CNN+LSTM+SA2017.09 | 76.6 | 64.4 | 82.7 | 77.4 | 82.6 | 75.4 | |
| CNN+LSTM2017.09 | 52.3 | 43.7 | 65.2 | 67.1 | 49.3 | 53 | |
| LSTM2017.09 | 46.8 | 41.7 | 61.1 | 69.8 | 36.8 | 51.8 | |
| Q-type baseline2017.09 | 41.8 | 34.6 | 50.2 | 51 | 36 | 51.3 |