Qualcomm Deep Learning Engineer interview questions

Introduce yourself Where else did I apply to? What is my plan after graduation? Ph.D vs Jobs What is my thesis about? What’s your own contribution? Feature based visual SLAM vs direct method What is the effect of illumination change on both types of SLAM? What are the benefits of using visual features such as ORB, SIFT etc. Explain epipolar geometry and homogeneous coordinate. Why SIFT is robust to illumination change, why does it use features pyramid Explain about monocular depth estimation Why do they use L1 loss in appearance matching? In packnet-sfm for example? Why does training a neural network can diverge sometimes? What can we do about it?