Image Enhancement for Pedestrian Detection at Night Time

Present days people are often geographically dispersed unevenly in low-light photographs, making it challenging to differentiate them due to the few photons and inherent noise. Some person detecting models are unable to function in dimly lit environments. Within this project First, we will use gamma correction, a method for improving low-light images and storing the improved photos. Second step is to utilizing the enhanced and low-light pictures, the convolution neural network technique is applied.The data will be used to train the CNN model, which produces results for accuracy, precision, recall, and f1. The SVM HOG features will be applied during prediction to extract people identification from low-light photographs. Our proposed model provides more prediction results compare to existing project.