Mitigating_Gender_Bias_In_Captioning_System
Image captioning bias study
An investigation into bias in image captioning systems using a dataset and a new model design to mitigate this bias
under review
13 stars
4 watching
2 forks
Language: Python
last commit: over 5 years agoLinked from 1 awesome list
Related projects:
| Repository | Description | Stars |
|---|---|---|
 lukemelas/image-paragraph-captioning | Trains image paragraph captioning models to generate diverse and accurate captions | 90 |
 visionjo/facerec-bias-bfw | This project provides a data proxy to evaluate bias in facial recognition systems across demographic groups. | 47 |
 i-gallegos/fair-llm-benchmark | Compiles bias evaluation datasets and provides access to original data sources for large language models | 115 |
 chapternewscu/image-captioning-with-semantic-attention | A deep learning model for generating image captions with semantic attention | 51 |
 tolga-b/debiaswe | Removes problematic gender bias from word embeddings to create more neutral representations. | 244 |
 contextualai/lens | Enhances language models to generate text based on visual descriptions of images | 352 |
 deeprnn/image_captioning | This implementation allows users to generate captions from images using a neural network model with visual attention. | 790 |
 nyu-mll/bbq | A dataset and benchmarking framework to evaluate the performance of question answering models on detecting and mitigating social biases. | 92 |
 privacytrustlab/bias_in_fl | This project investigates how bias can be introduced and spread in machine learning models during federated learning, and aims to detect and mitigate this issue. | 11 |
 jiasenlu/adaptiveattention | Adaptive attention mechanism for image captioning using visual sentinels | 335 |
 dssg/aequitas | Toolkit to audit and mitigate biases in machine learning models | 701 |
 ethicalml/xai | An eXplainability toolbox for machine learning that enables data analysis and model evaluation to mitigate biases and improve performance | 1,135 |
 rudinger/winogender-schemas | A dataset of manually crafted sentence templates to evaluate the presence of gender bias in natural language processing systems | 70 |
 dvlab-research/lisa | A system that uses large language models to generate segmentation masks for images based on complex queries and world knowledge. | 1,923 |
 dib-lab/2012-paper-diginorm | Provides source code and analysis pipeline for evaluating an image normalization technique using publicly available data | 8 |