[Proposal]: Computer Vision in Self-Driving Cars

[Ridhwanluthra]

• Abstract (2-3 lines) Self driving cars are an exciting area of research for all data enthusiasts at this time. Getting a machine to process the richness of the real world and make ethical decisions. A major part of this is using camera images to extract information from the world. This talk is focused on allowing cars to perceive the world the way humans would. The aim here is to discuss the state of the art.
• Brief Description and Contents to be covered

  Computer Vision in Self Driving Cars

  Image Processing

• Basics of image processing.
• Geometric transforms on images.

Convolutional Neural Networks

• Working of CNNs from scratch.
• Common architectures.
• Intuitions on developing your own architectures.

Behavioral Cloning

• Running Udacity's open source simulator.
• Data augmentation
• Training to replicate human behavior.

---

• Pre-requisites for the talk None
• Time required for the talk 40 mins
• Link to slides https://magic.piktochart.com/output/23596250-self-driving-car-workshop
• Will you be doing hands-on demo as well? can be, in that case, time will be increased to 2 hrs
• Link to ipython notebook (if any) will update
• About yourself I am a Robotics and AI enthusiast currently pursuing B.tech in information technology and mathematical innovations at Cluster Innovation Centre. I work on building robots with advanced AIs to solve real-world problems.
• Are you comfortable if the talk is recorded and uploaded to PyData Delhi's YouTube channel? yes
• Any query? no

[shagunsodhani]

May be consider doing the talk in 2 parts: Conv and then behavior cloning. The slides for cloning part are sparse - I assume that is because this is a WIP.

[Ridhwanluthra]

no that's because that part mainly is hands-on and relies on a jupyter notebook, no slides. so if it's a talk behaviour cloning will be covered in brief (basically a demo). if hands-on then will do in depth.

[shagunsodhani]

Then lets wait for the notebook :)

[manojpandey]

@Ridhwanluthra ping - do you plan to give this talk ?