Culture-mixup

To Do List

• [✓] Upload Cultural Concepts Dataset

  • Status: Completed
  • Description: The dataset, featuring cultural concepts from five different countries and regions, has been successfully uploaded. It provides detailed explanations and corresponding images.
  • Dataset Link: https://huggingface.co/datasets/zhili312/multimodal-cultural-concepts

• [✓] Upload Cultural Concepts Adaptation Dataset

  • Status: Completed
  • Description: This dataset includes both the semantic relationships based on ConceptNet from the original paper and fuzzy searches conducted by ChatGPT.

• [✓] Upload Cultural Concepts Adaptation Code

  • Status: Completed
  • Description: The provided code can be used to replicate the above Cultural Concepts Adaptation Dataset and can also be applied to create general cultural concept adaptation datasets.
  • Usage Instructions:
    1. Install the necessary dependencies listed in requirements.txt.
    2. Download the Cultural Concepts Dataset.
    3. For example, to use Chinese as the source language, run the following command:

       python Buildgraph.py --source_language zh --target_language en --input_file Chinese/Chinese.xlsx --output_file Chinese_English_adaptation_with_chatgpt.xlsx --use_chatgpt
       

• Upload Culture-mix Code

  • Status: Coming Soon

• Upload the Latest Multimodal Language Model (GPT-4-V and Minigpt-4) Test Dataset and Evaluation Results

  • Status: Coming Soon