Case Study: How to Make a Really Attention-Getting Demo

Back in August 2011, Jeff Bier's editorial "How to Make a Really Annoying Demo" neatly summarized the common attributes of poorly developed and executed product demonstrations that he'd auditioned over the years. Recently, Bier (and company) had the opportunity to show the ability to "practice what is preached," in the context of a demo developed in partnership with Analog Devices for that company's newly introduced BF60x Blackfin SoCs (see "Analog Devices' Latest Blackfin Proliferations Get Embedded Vision Religion" in this month's edition of InsideDSP).

Two of the four members of the BF60x family embed a specialized processing core called the PVP (pipelined video processor), custom-tailored for automotive driver assistance and other embedded vision applications. Perhaps obviously, therefore, it was appropriate for the PVP's capabilities to receive prominent showcase in the demo, along with those of the dual Blackfin DSP cores, the abundant on-chip memory and peripherals, and other features. But what particular system implementation would best accomplish Analog Devices objective of quickly and clearly communicating the chips' attributes both versus competitors' offerings and Blackfin predecessors?

Embedded vision, succinctly defined, refers to systems that understand their environments through visual means. It involves not only capturing, archiving and playing back still images and video, as is done in photography and videoconferencing, but using digital processing and intelligent algorithms to interpret meaning from (and appropriately respond to) this captured content. Applications such as optical character recognition and road sign detection and identification would certainly fit the description, but demonstrations of these capabilities wouldn't be obviously superior (speed, reliability, etc) to the human visual system and brain's capabilities, thereby leading to the potential for a less-than-stellar demo evaluation by onlookers.

After much brainstorming, Analog Devices came up with the idea of a dice-counting demonstration. BDTI planted that idea seed and grew a comprehensive design out of it. A video camera, connected to a Blackfin evaluation board, captures images of seven or more dice tossed below it. The PVP, in combination with the Blackfin DSP cores, discerns the displayed value of each die face, and then sums the individual values into a total which is displayed on a video monitor alongside a picture of the splayed dice. BDTI's engineers came up with several embedded vision algorithm alternatives, befitting the multi-function nature of the PVP. And Analog Devices and BDTI jointly selected the final implementation after evaluating the multiple algorithm candidates for speed, accuracy, reliability and other factors.

Did the Blackfin demo successfully avoid the "four time-tested techniques that are virtually guaranteed to lead to a crummy demo," that Bier wrote about in August 2011? Let's see:

1. Lack clarity regarding your goal: This demo's primary purpose, as Jeff described in his editorial, was "to attract a crowd to a trade show booth" (along with attracting the attention of press and analyst representatives during product briefings). In this respect, it succeeded handily, as Dean Takahashi's (VentureBeat) and Junko Yoshida's (EE Times) subsequent coverage exemplifies. And it was also an effective attention-getter at the March 2012 Embedded Vision Alliance Summit, where Analog Devices demonstrated it to attendees.
2. Make it complicated: Rare is the individual who's never played a game that involves tossing and then counting up dice. As such, although the means by which the demo accomplished its objective was complex, it required no in-depth explanation for onlookers. As Jeff wrote, "the demo speaks for itself: that is, just by watching the demo, the audience immediately understands what the demo is doing and why it's worth watching."
3. Make it too simple: Had Analog Devices and BDTI gone with one of the alternative embedded vision demo ideas they considered, this potential pitfall might have been a problem. Human beings know what stop signs look like, and most of them can read text. That a computer can do the same thing is not terribly impressive. But that it can discern and count dice orders of magnitude faster than a human being...that's memorable.
4. Lie: By virtue of the fact that Analog Devices and BDTI, both longstanding and reputable companies, were involved in the project, this potential issue wasn't even on the radar screen. The partners were careful to not "lead the audience to believe that the product is more capable, robust, or general than it really is", because after all, "sooner or later, the prospective customer will discover the truth. And a prospective customer who feels duped is very unlikely to ever become an actual customer."

BDTI can help you craft a compelling demo to effectively communicate the capabilities of your next product. Learn how by contacting Jeremy Giddings at +1 925 954 1411 or giddings@BDTI.com.