Images of Sylvester Stallone used to train models for recognizing stroke
A dataset on Kaggle intended for training stroke detection models has been criticized for including images of celebrities like Sylvester Stallone. Researchers Adrian Barnett and Alexander Gibson have highlighted the dataset's numerous duplications and ethical concerns regarding its use in clinical research. This incident reflects a broader issue of questionable data quality in scientific literature derived from online repositories.
- ▪The dataset in question includes images of celebrities and has been labeled as 'comically bad' by a statistician.
- ▪It was used to train a predictive clinical model for early stroke detection, raising ethical and scientific concerns.
- ▪The researchers found that many images in the dataset depicted Bell's palsy and children, rather than actual stroke patients.
Opening excerpt (first ~120 words) tap to expand
‘Comically bad’ datasets used to train clinical models for stroke and diabetes A dataset on Kaggle purportedly showing people who have had a stroke includes images of Sylvester Stallone from Rambo and other celebrities. Source Scrolling through an online image dataset, Adrian Barnett, a statistician at the Queensland University of Technology in Australia, pointed out a few familiar faces. Sylvester Stallone as Rambo, and then again on the red carpet. “This is just ridiculous,” Barnett said. George Clooney, Angelina Jolie and Daniel Craig all appear more than once, often with the same image.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Retraction Watch.
Discussion
0 commentsMore from Retraction Watch
