QUESTION 7 [15 MARKS] You are the Chief Al Engineer for an autonomous vehicle company, You want to use an object detection model for the self-driving vehicle to detect road signs, pedestrians and other vehicles. (a) Explain your plan according to the data driven machine learning approach, which consists of three main steps: (1) dataset preparation, (2) training/optimization of the object detection model, and (3) model testing/implementation. Describe each step in detail. Recommend a hardware for training and another for inferencing (if different). [9 marks) (b) For this specific application, discuss which is more important: inference accuracy or speed? [3 marks) (c) Recommend an object detection model by selecting an architecture as well as a backbone network. You may refer to the graph below, or recommend other models. Justify and explain your selection. [3 marks) 40 Faster CNN w/Reset Ret, Proposal Meta Architecture Faster RCNN R-FON SSD 35 FN! Reset. Mies 100 Proposal 30 Paster RCNN w/caption Reet, Mes 100 Prepel, Stride Overall mAP 25 20 Feature Extractor Inception Resnet V2 Inception V2 Inception V3 MobileNet Resnet 101 VGG SSD w/Inception V2, to Rus SSD w/Mobilet, totes 15 10 0 200 400 600 800 1000 GPU Time