Visual Question Answering on CLEVR-Humans
99.8AccuracyNS-VQA
Evaluation Results
| Method | Links | |
|---|---|---|
| NS-VQAExtra supervision=Programs, Evaluation Protocol=Finetuned2024.11 | 99.8 | |
| IPRMExtra supervision=None, Evaluation Protocol=Finetuned2024.11 | 85.5 | |
| MDETRExtra supervision=Bound. Box, Evaluation Protocol=Finetuned2024.11 | 81.7 | |
| MACfine-tuned=true2018.03 | 81.5 | |
| MACExtra supervision=None, Evaluation Protocol=Finetuned2024.11 | 81.5 | |
| CNN+GRU+FiLMfine-tuned=true2018.03 | 75.9 | |
| FiLMExtra supervision=None, Evaluation Protocol=Finetuned2024.11 | 75.9 | |
| NS-VQAExtra supervision=Programs, Evaluation Protocol=Zero-shot2024.11 | 67.8 | |
| PG+EEextra supervisory information=true, Program Labels=18K, fine-tuned=true2018.03 | 66.6 | |
| PG+EEExtra supervision=Programs, Evaluation Protocol=Finetuned2024.11 | 66.6 | |
| IPRMExtra supervision=None, Evaluation Protocol=Zero-shot2024.11 | 63.8 | |
| MDETRExtra supervision=Bound. Box, Evaluation Protocol=Zero-shot2024.11 | 59.9 | |
| CNN+LSTM+SA+MLPfine-tuned=true2018.03 | 57.6 | |
| MACfine-tuned=false2018.03 | 57.4 | |
| MACExtra supervision=None, Evaluation Protocol=Zero-shot2024.11 | 57.4 | |
| CNN+GRU+FiLMfine-tuned=false2018.03 | 56.6 | |
| FiLMExtra supervision=None, Evaluation Protocol=Zero-shot2024.11 | 56.6 | |
| PG+EEextra supervisory information=true, Program Labels=18K, fine-tuned=false2018.03 | 54 | |
| PG+EEExtra supervision=Programs, Evaluation Protocol=Zero-shot2024.11 | 54 | |
| CNN+LSTM+SA+MLPfine-tuned=false2018.03 | 50.4 | |
| CNN+LSTMfine-tuned=true2018.03 | 43.2 | |
| CNN+LSTMfine-tuned=false2018.03 | 37.7 | |
| LSTMfine-tuned=true2018.03 | 36.5 | |
| LSTMfine-tuned=false2018.03 | 27.5 |