/v2

W251 2018 reload

Primary LanguageJupyter Notebook

W251 reloaded (Fall 2019)

The revised class is focused on Deep Learning and Big Data at the Edge and in the Cloud.

To follow the class, you'll a Mac or a PC (Windows or Linux) with an ability to run docker or VirtualBox VMs. You will also need additional equipment as follows:

Required equipment:

  1. Nvidia Jetson TX2 dev kit ($299 + tax with educational discount). To get the discount, you'll need a code, which you can get from Nvidia directly or by signing up for the class or contacting your instructor(s). We are currently testing Nvidia Xavier and you're welcome to give it a shot, but it's expensive ($899 AFTER the educational discount), and we are not actively testing the class materials on it. You've been warned.

  2. Additional storage. The Jetson TX2 System on a Chip (SoC) has only 16G eMMC, which is tight for developer. There is a slot for an SD card and it could be the cheapest (albeit, a somewhat slow) option. The class will take advantage of Docker which in turn drive up storage needs. We recommend at least 128 GB, for instance, this card (currently, $19.99). If you are a performance user, you will need to get a SATA SSD, e.g. this one (currently, $35.50). While it's possible to insert an SSD directly / vertically into the Jetson dev board, it's quite easy to accidentally bump the SSD causing the slot to break off the board. Therefore, it's safer to purchase a cable connector (currently, $5.66). Alternatively, you could just use any USB 3.0 external disk.

  3. A external webcam. The TX2 does have an onboard camera, but it has always had compatibility issues. So a USB webcam is a safe option, e.g. this cheap camera. Note that this particular webcam is not HD, even though it's advertised as such and you may want to go with a true HD webcam instead. But, true HD is not required for the class.

  4. A USB 3.0 switch . Here's the one I am happy with (currently, $9.89) If you are using Xavier, you'll need this one instead

  5. You will also need a mouse, keyboard, a monitor and an HDMI cable to connect to the monitor. The tx2 has wifi built in or a wired connection is preferred if you'll be doing a lot of video processing.

  6. While we have not had any issues, an ant-static wrist strap is never a bad idea.

Homeworks:

A homework is due before each class. There are two types of homeworks: graded and credit only. Here is the link to class 1 - be sure to complete the setup of your Jetson as described in homework 1

SSH Reminder:

Ensure that any VSI/VM create prohibts login with password prohibited. See: https://github.com/MIDS-scaling-up/v2/tree/master/week02/hw/README.md for details.

Graded homeworks

The graded homeworks are week 3, week 6, week 9, week 11; please notice that those are the slots that are available on the ISVC website except there are labeled as Homework 1, 2, 3 and 4 (we are working on getting this changed but in the meantime submit matching the graded homework week with it's slot i.e week 3 uploaded to Homework 1).

Final Projects (due in the final class session of the semester)

  • Form teams of three to four people
  • Leverage big data, cloud, DL, and the edge device to do something cool
  • Should be more than you can do on a workstation
  • To turn in: White paper explaining what you did, how you did it, what you learned, what went right (or wrong) and a brief presentation (10-15 minutes)

An example of a final project:

  • Leverage a dataset of missing persons.
  • Train a model in the cloud to recognize the missing persons.
  • Deploy the model to your edge device.
  • If a person is recognized, send the image and location back to the cloud for further actions.

Softlayer