Script for Composed Image Retrieval (CIR) Dataset Generation using GPT-4o
Description
Create a user-friendly script that enables the generation of a Composed Image Retrieval (CIR) dataset leveraging the GPT-4o model. The script should encapsulate the full pipeline:
Requirements
The script should:
- Allow users to:
- Specify their input dataset.
- Provide their OpenAI API key.
- Customize their prompts for GPT-4o.
- Implement all three stages of the CIR pipeline.
- Include post-processing logic to clean, validate, and store the final dataset.
- Be modular but runnable end-to-end from a single script.
- Include in-line comments directing users on:
- Where to specify their dataset.
- How to insert their API key.
- What the output directory will be.
- Where to specify custom prompts.
Deliverables
- A single script:
generate_dataset.py
- A
requirements.txt file listing all necessary dependencies.
Script for Composed Image Retrieval (CIR) Dataset Generation using GPT-4o
Description
Create a user-friendly script that enables the generation of a Composed Image Retrieval (CIR) dataset leveraging the GPT-4o model. The script should encapsulate the full pipeline:
Requirements
The script should:
Deliverables
generate_dataset.pyrequirements.txtfile listing all necessary dependencies.