I2I - From illusion to illumination: A neoteric deep learning model for recognizing medical situation actions using...

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Robot vision systems frequently encounter illumination challenges.
• Existing human action recognition methods inadequately address illumination variations.
• Poor illumination significantly impacts the accuracy of medical action recognition, potentially leading to critical errors.

Purpose of the Study:

• To develop a robust deep learning model for accurate medical action recognition despite varying illumination.
• To address the sensitivity of robot vision to lighting conditions in medical contexts.
• To introduce the Illusion to Illumination (I2I) model for enhanced performance.

Main Methods:

• A deep learning-based model named I2I (Illusion to Illumination) was proposed.
• Depth data from the NTU RGB+D dataset was utilized.
• Features were extracted using Histogram of Depth (HoD) and refined with a threshold mechanism.

Main Results:

• The I2I model demonstrated effective identification of medical actions in low-light environments.
• The model achieved a recognition accuracy of 91.15%.
• Performance was validated against state-of-the-art methods, showing superiority.

Conclusions:

• The I2I model significantly improves medical action recognition accuracy under adverse illumination.
• The proposed approach offers a reliable solution for vision-based medical tasks in robotics.
• The study highlights the importance of addressing illumination challenges in robot vision for critical applications.