35th Annual Student Research Conference
Implementing Convolutional Neural Networks to Identify Pedestrians
Justin A. Caringal
Dr. Ruthie Halma, Faculty Mentor
Accurate identification of pedestrians is an important task required in many computer vision and image processing problems. The increasing ubiquity of self-driving cars and the need for image processing in other application areas prompted our study of the efficacy of producing an artificial intelligence model using only publicly available resources and documentation. This paper discusses the process undertaken to create a convolutional neural network model capable of distinguishing between people and landscapes with a 100 percent validation accuracy at a 0.16 percent validation loss. Although this model is not ready for commercial use in self-driving cars, it provides a good starting point for additional improvements. This paper presents a final model, the culmination of the work done during this project; however, it is by no means a final version. Potential improvements are discussed.
Keywords: Artificial Intelligence, Neural Networks, Convolution, Datasets, Deep Q-Learning
Topic(s):Computer Science
Presentation Type: Oral Presentation
Session: 107-2
Location: MG 1098
Time: 8:45