A deep transfer learning based convolution neural network framework for air temperature classification using human clothing images
2024

Classifying Air Temperature from Clothing Images Using Deep Learning

Sample size: 10000 publication 10 minutes Evidence: high

Author Information

Author(s): Ahmed Maqsood, Zhang Xiang, Shen Yonglin, Ali Nafees, Flah Aymen, Kanan Mohammad, Alsharef Mohammad, Ghoneim Sherif S. M.

Primary Institution: China University of Geosciences, Wuhan, China

Hypothesis

Can deep learning techniques accurately classify air temperature levels based on human clothing images?

Conclusion

The study demonstrates that a deep transfer learning framework can effectively classify air temperature levels from clothing images, achieving high accuracy.

Supporting Evidence

  • DenseNet121 achieved the highest accuracy of 98.13%.
  • The study utilized a dataset of 10,000 images categorized into high and low temperature classes.
  • Semantic segmentation improved model performance significantly.
  • Grad-CAM was used to visualize important features in the classification process.
  • Models were evaluated using metrics like accuracy, precision, and recall.
  • 80% of the dataset was used for training and 20% for testing.
  • Deep learning methods outperformed traditional machine learning techniques in this context.
  • The study highlights the potential for practical applications in weather prediction and environmental monitoring.

Takeaway

This study shows that we can tell if it's hot or cold outside just by looking at what people are wearing in pictures.

Methodology

The study used a deep transfer learning framework with various CNN models to classify air temperature levels from a dataset of clothing images.

Potential Biases

Potential bias may arise from the dataset being limited to specific demographics or clothing styles.

Limitations

The study may be limited by the dataset size and the specific types of clothing images used.

Participant Demographics

The dataset included images from 200 students, comprising undergraduates, graduates, and postgraduates.

Statistical Information

P-Value

p<0.05

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1038/s41598-024-80657-y

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication