|
Abstract:
|
With large amounts of data being sent and received every day, computers need faster working algorithms to analyze data more efficiently. This poster analyzes and describes, both qualitatively and quantitatively, the current bias among popular object detection data sets. For example, bias in an image data set would be all the objects being the same, or all the images occurring in the night. Furthermore, this poster describes the impact of data set bias on the learned features of deep learning models. This is done by using several techniques such as algorithm analysis, statistical inference and modeling. Purdue University's CAM2 (Continuous Analysis of Many Cameras) team is led by Dr. Yung-Hsiang Lu. This material is based upon work supported by the National Science Foundation under Grant No. 1246818.
|
