Recognizing human actions in dark-light conditions has many real-time applications like night-smart surveillance systems, elderly people, monitoring in smart homes, military applications, self-driving cars, etc. The current state-of-the-art techniques cannot effectively perform human action recognition in dark light situations. In the proposed work, we have constructed a novel deep learning architecture that involves an image enhancement module followed by an action classification network, to classify the actions in dark/low-light videos. Our approach surpasses the available state-of-the-art techniques for action recognition in the dark and provides 96.60% Top 1 accuracy.

Several research works have been developed in the area of action recognition. Unfortunately, when these algorithms are applied to low-light or dark videos, their performances are highly affected and found to be very poor or fall rapidly. To address the issue of improving the performance of action recognition in dark or low-light videos; in this article, we have developed an efficient deep 3-D convolutional neural network based action recognition model. The proposed algorithm follows two-stages for action recognition. In the first stage, the low-light videos are enhanced using zero-reference deep curve estimation, followed by the min–max sampling algorithm. In the latter stage, we propose an action classification network to recognize the actions in the enhanced videos. In the proposed action classification network, we explored the capabilities of the $R(2+1)D$ for spatio-temporal feature extraction. The model's overall generalization performance depends on how well it can capture long-r...