👉 The original dataset **HaVQA* is available on: Hugging Face - HaVQA.
The **HaVQA_aug** dataset was constructed from the original dataset **HaVQA** through offline data augmentation.
The table below presents the statistics of this dataset.
| Jeu de données | Train | Test | Images |
|--------------------------|--------|-------|--------|
| HaVQA_aug | 9 625 | 2 407 | 3 110 |
The authors of the original HaVQA dataset are:
> Shantipriya Parida, Idris Abdulmumin, Shamsuddeen Hassan Muhammad, Aneesh Bose, Guneet Singh Kohli, Ibrahim Said Ahmad, Ketan Kotwal, Sayan Deb Sarkar, Ondřej Bojar, Habeebah Kakudi
> *HaVQA: Hausa Visual Question Answering Dataset*.
> Lien vers l'article
| **LLMs** | **Image encoder** | **WuPalmer** | **Accuracy** | **F1-score** |
|-----------|-------------------|--------------|---------------|--------------|
| mT0-base | vit-base-patch16-224-in21k | 15.03% | 15.03% | 0.38% |
| mT0-base | clip-vit-base-patch32 | 14.44% | 14.49% | 0.35% |
| mT0-base | mae-base | 15.37% | 15.38% | 0.37% |
| mT0-base | deit-base-distilled-224 | 15.54% | 15.54% | 0.37% |
| mT0-large | clip-vit-base-patch32 | 15.61% | 15.61% | 0.41% |
| mT0-large | deit-base-distilled-224 | 15.45% | 15.45% | 0.37% |
| afriberta-large | vit-base-patch16-224 | 17.29% | 17.19% | 0.57% |
| afriberta-large | clip-vit-base-patch32 | 17.02% | 17.19% | 0.57% |
| afriberta-large | mae-base | 18.27% | 18.27% | 0.71% |
| afris-bert-large-76L | vit-base-patch16-224 | 17.26% | 16.96% | 0.59% |
| afris-bert-large-76L | clip-vit-base-patch32 | 17.36% | 16.96% | 0.59% |
| afris-bert-large-76L | mae-base | 18.12% | 18.11% | 0.67% |
| gemini | vit-base-patch16-224-in21k | 14.70% | 14.50% | 0.36% |
| gemini | clip-vit-base-patch32 | 14.04% | 14.04% | 0.31% |
| bloomz560 | vit-base-patch16-224-in21k | 17.09% | 17.09% | 0.55% |
| bloomz560 | mae-base | 17.62% | 17.62% | 0.59% |
| bloomz7b1 | vit-base-patch16-224 | 17.36% | 17.36% | 0.59% |
| bloomz7b1 | mae-base | 17.62% | 17. …