Thanks for great structured course and valuable learning materials. I’m happy to be part of this learning community. I’ve finished part 1 and happy to learn part 2 online.
Lately I was looking for DL job and I’ve got one interview proposal. I live in Tel-Aviv and I hope that there is enough relevant positions. Tomorrow, (in 24 hours) I’m going on job interview and over the phone I was asked to detect football ball and track it in video using keras. This sound like standard task of Lesson 8, but I would love to get tips and hints related to this topic.
I’m going over CS231n lectures and visiting local meet ups, learning about different approaches to solve vision tasks.
I read this post, and found it useful and interesting.
Nevertheless I would like to ask “what technical questions would you expect on this kind of job”? Thanks for fast answers, they probably will have big influence on how my interview goes.
Thanks a lot!
I had a 3 and a half hour interview and I started to work on 1st of April, Sunday (in Israel week starts on Sunday, but Friday and Saturday is day-offs).
I didn’t prepare to interview in special way - just was learning fast.ai lessons with great pleasure.
I was going to sing NDA about job, but it will be video processing using OpenCV (not ML part) and using Mask R-CNN in production (most interesting and challenging part). If you have any suggestion about good accuracy models that do segmentation really fast - I’d be happy to learn about them.
I was honest and humble about my knowledge in DL, as I’m learning it for only 6 months. Instead I got many question not connected with technology or coding itself.
First task was to write program that finds longest palindrome in given string. I had visual studio and wrote it within 40 or 50 minutes in C# (cause I worked with it for ~3 years).
Next questions were about math and physics - how to find trajectory of falling ball from the picture (2d) with blurred ball. Correct answer was to take shutter time from image metadata, calculate horizontal velocity from known time and distance in picture. Addition to this question was how to calculate it in 3d.
There were bunch of question about vector calculation and finding angles - many basic stuff.
Part that I really loved was logical questions, or how I like to call them “out of the box thinking”. I’m attaching picture with the task.
Funny part - the interview I had was with remote call and desktop sharing, so technical lead asked me questions by painting in Microsoft Paint and I was answering the same way.
From left to right: First square is divided into 2 equal parts, second is divided into 4 equal parts. Task - divide last square into 5 equal parts.
I’m really happy that I’m working in field of Deep learning. This course gives valuable knowledge, but even more important for me - I got self confidence (in positive meaning) in overcoming tasks, understanding black boxes and working in this field.