Search

This work aims to introduce a novel approach for auxiliary task guidance (ATG). In this approach, our goal is to achieve effective guidance from a suitable auxiliary task by utilizing the uncertainty in calculated gradients for a mini-batch of samples. Our method calculates a probabilistic fitness factor of the auxiliary task gradient for each of the shared weights to guide the main task at every training step of mini-batch gradient descent. We have shown that this proposed factor incorporates task specific confidence of learning to manipulate ATG in an effective manner. For studying the potency of the method, monocular visual odometry (VO) has been chosen as an application. Substantial experiments have been done on the KITTI VO dataset for solving monocular VO with a simple convolutional neural network (CNN) architecture. Corresponding results show that our ATG method significantly boosts the performance of supervised learning for VO. It also out performs state-of-the-art (SOTA) auxiliary guided methods we applied for VO. The proposed method is able to achieve decent scores (in some cases competitive)compared to existing SOTA supervised monocular VO algorithms, while keeping an exceptionally low parameter space in supervised regime.

A growing body of research has applied intelligent transportation technologies to reduce traffic congestion at signalized intersections. However, most of these studies have not considered the systematic integration of traffic data collection methods when simulating optimum signal timing. The present study developed a three-part system to create optimized variable signal timing profiles for a congested intersection in Dhaka, regulated by fixed-time traffic signals. Video footage of traffic from the studied intersection was analyzed using a computer vision tool that extracted traffic flow data. The data underwent a further data-mining process, resulting in greater than 90% data accuracy. The final data set was then analyzed by a local traffic expert. Two hybrid scenarios based on the data and the expert’s input were created and simulated at the micro level. The resultant, custom, variable timing profiles for the traffic signals yielded a 40% reduction in vehicle queue length, increases in average travel speed, and a significant overall reduction in traffic congestion.