Hand Gesture Recognition in Low Light Conditions

This project addressed the challenge of hand gesture recognition in lowlight conditions, where traditional systems often fail due to poor visibility and degraded image quality.We developed a gesture recognition model capable of detecting three specific gestures-palm open, swipe right, and thumbs up-under varying illumination levels.The system utilized image enhancement techniques, region of interest (RoI) segmentation, and a Convolutional Neural Network (CNN) for accurate gesture classification.We trained the model on a custom dataset that included diverse lighting scenarios and evaluated its performance by tracking accuracy and loss over multiple training epochs.The model showed consistent improvements in recognition accuracy and convergence behaviour, indicating effective learning and robustness against lighting variations.Although the system was limited to a small set of gestures, it demonstrated reliable performance in controlled low-light environments.This work contributes to the ongoing development of accessible and adaptive gesture-based interfaces, with potential for future expansion to more complex gesture sets and broader real-world use cases.