Member Info for johnchucka

Semicast Research@semicast_res

.@seeingmachines G2M strategy is approx 4 years ahead of market understanding. Capacity limitation means rising prices, not forced takeover. Compare with airlines, hotels, Uber surge pricing etc. Fovio strat. mostly avoids capacity limitation. V. smart thinking here.

Semicast Research@semicast_res

Trading update from @seeingmachines also confirms Fovio at heart of multi-sector strat. Same tech for key sectors (auto & Guardian Gen3). Fovio platform is massively scalable. V. smart Go-to-market strat. & great use of eng. resource. SM not obviously capacity limited.

Semicast Research@semicast_res

V. approx calcs share holdings
@seeingmachines 12.2%
VSI 11.2
Miton 7.0
L&G 4.9
Ind IT 4.5
Herald 4.1
Lombard 3.7
FIL 3.6
Royce 1.9
Treeptop 1.5
Axa 1.4
Polar 1.4
City 0.8
Hargreave Hale 0.7
Total >50.1%
Some data old but affirms strategy of UK chairman with CEO in Canberra.

Semicast Research@semicast_res

ONLY FOR FUN:

NVDA marcap 13 Mar 17: $65B
NVDA marcap 02 Aug 18: $151B

MBLY purchase price 13 Mar 17: $15.3B
MBLY purchase price 02 Aug 18: ~$35B ?

http://eetimes.jp/ee/articles/1808/03/news038.html

To improve the experience of riding a next-generation car, to understand human emotions Automotive AI

On August 1, 2018, CAC (CAC) announced that it will offer a solution "Automotive AI" that analyzes the emotions of automobile drivers and passengers in real time from the same day.
The same solution was developed by Affectiva of a US venture company. Sensing facial expression data and voice data of passengers with cameras and microphones mounted in the car and analyzing feelings by deep learning based on them.

Automotive AI is developed as emotion recognition AI (artificial intelligence) optimized for use in automobiles, and it is assumed to be "Technology to realize a new passenger experience for cars" (CACY). There are four emotional values such as joy, anger, surprise, Valence (positive expressions / negative expressions), eight emotional values such as smiling faces, raising the eyebrows, wrinkling between the eyebrows, indexes concerning three drowsiness, face orientation and Angle, multimodal analysis using three speech emotions as indicators.

Therefore, it is possible to specify the complicated condition of the driver such as drowsiness and attentiveness lowering, and it is said that the possibility of preventing accidents is increased compared with the conventional system that depends on head position and eye gaze measurement. Also, it is assumed that the effect can be expected for commercial use of automatic driving.
Moreover, the solution does not require an external network connection for operation, it features that it can be used in the local processing mode completed by the in-vehicle system. We list quad core Arm64 processors (2.15 GHz) or dual core x86 processors (2.3 GHz) and 350 Mbytes of memory as required, and operate on Ubuntu Linux 16.04. Since both visible light and near infrared cameras are available, it is assumed that driver's expression can be correctly detected even in backlighting and darkness.

Because it utilizes deep learning, identification accuracy is improved by collecting and analyzing human expression data. The developer, Affectiva, said cooperation with OEM (automaker) such as DAIMLER and BMW, semiconductor makers such as Intel and NVIDIA, etc. are progressing development.
The solution is planned not only to be installed in private cars but also to be used as a solution for drivers of public transportation such as trains, buses and taxis.

Semicast Research@semicast_res

Key points @seeingmachines y/e update
"The addition of further models is currently in discussion" OEMs introduce new tech in a few models first. That is a well proven strategy i.e. MBLY
"Automotive business now has established offices in Japan" SM & SE both in Japan for a reason!

Semicast Research@semicast_res

Semicast ADAS&AD 2018 Report

Auto Front Camera F/cast:
2015 9.3M
2021 39.9M
2024 58.3M
CAGR 22%
Leader:MBLY(bought by Intel US$15.3B)

Auto Camera DMS F/cast:
2015 0.25M
2021 21.3M
2024 53.7M
CAGR 81%
Leaders: @seeingmachines (marcap US$320M); @smarteyeAB (marcap US$55M)

Semicast Research@semicast_res

Lots of great research on DMS can also be found on the SEE chat room on http://www.lse.co.uk , where a group of private investors digest issues and events related to @seeingmachines. Users Seeing2020 and Lewbo18 are especially knowledgeable.

Nissan Seamless Autonomous Mobility - Maarten Sierhuis

https://www.youtube.com/watch?v=MK0u-89PPho

Semicast Research@semicast_res

Very rough guide only.
Xilinx Zynq-7020 ~2W
Mbly EyeQ4 ~3W
Intel Atom E620 ~3W
Nvidia Drive PX2 ~250W
Can't find power info for i.MX8 family yet.

Shows how far nvidia solutions are from other embedded processing elements typical in automotive. Serious heat issues at 250W TDP.

Semicast Research@semicast_res

Next DW for @SmartEyeAB or @seeingmachines ? The action seems to have moved to Chinese OEMs all of a sudden.

http://www.thedrive.com/tech/21857/chinese-electric-car-startup-nio-delivers-first-es8-crossovers
……...
Xilinx Zynq-7000 and NXP i.MX8 are both 28nm process geometry. For a given wattage, @seeingmachines will have higher DMS performance than @SmartEyeAB owing to power optimization offered by customization in FPGA fabric. Mbly EyeQ3 is 40nm (old); EyeQ4 is 28nm & enters prod. 2018.

FYI..The JC 11.24 post was Semicast Tweet.. apologies forgot to put Semicast reference.

FPGAs are one of the most complicated processing elements to understand, hence why rarely talked about. Here's a great article on FPGA in automotive from the always excellent @junkoyoshida

https://www.eetimes.com/document.asp?doc_id=1333419

Semicast Research@semicast_res

Financial analysts seem oblivious to how sophisticated the Fovio platform from @seeingmachines is. Multi-sector strategy took a lot of cash to develop but FPGA is perfect fit. Upgrade path to Xilinx Zynq UltraScale+ MPSoC from Zynq-7000 future proofs the investment. v.v. smart.
.

@SmartEyeAB developing DMS on i.MX8 MCU from NXP (former product line from Freescale) and appear to have a hard DMS IP stack to run on Nvidia GPU. @seeingmachines Fovio is based on Xilinx Zynq-7000 FPGA. FPGA offers customization in both logic (HW) and SW; MCU is SW only.

The genius of @seeingmachines Fovio is that every auto OEM could sign up for SoP 2021. The platform is massively scalable with a SoTA DM engine coded in the FPGA fabric. OEMs & T1s can customize in time. Hard eng. work by SM mostly over for auto, now need more execution staff.

Some of the Recent news from Toyota does give me the impression they are playing catch up. No doubt I'm biased.

https://www.denso.com/global/en/news/news-releases/2018/20180302-g01/

Tokyo, Japan, March 2, 2018—Toyota will establish a new company in Tokyo in the latter part of this month named “Toyota Research Institute-Advanced Development” (TRI-AD) that will accelerate its efforts in advanced development for automated driving.
To enable the new efforts at TRI-AD, Toyota Motor Corporation (TMC), Aisin Seiki Co., Ltd. (Aisin), and DENSO Corporation (DENSO) have concluded a memorandum of understanding on joint development of fully-integrated, production-quality software for automated driving. Going forward, the three companies will hold further discussions, aiming to conclude a concrete joint development contract.
Together, TMC, Aisin and DENSO plan to invest more than 300 billion yen in TRI-AD. The new company is targeting a staff of approximately 1,000 employees, including external recruitment and staff from TMC, TRI, and Toyota Group Companies Aisin and DENSO. Toyota is in the process of selecting a location in Tokyo that is competitive in terms of accessibility and recruitment.

Semicast Research@semicast_res

A possible scenario: Honda agrees an all models DW with @Veoneer1 and @seeingmachines for SoP 2021. This puts tremendous pressure on Toyota in the mass market, esp. in Japan, EU and US, also Mazda, Subaru etc. thus starting a domino effect. Events could go VERY fast from here.

Semicast Research@semicast_res

2017 ADAS vendor market shares, with estimated camera DMS partnerships:
Conti/SM 19%
Bosch/SM 17%
Veoneer/SM 12%
Aisin-Denso/poss. FN or in-house 12%
ZF/SE 9%
Aptiv/SE 8%
Magna/SM 7%
Valeo/SE 4%
Others 12%

@seeingmachines look to be partnered with top 3 ADAS vendors

Continental courts Japan as upheaval hits industry

April 14, 2018

http://www.autonews.com/article/20180414/OEM10/180419920/continental-courts-japan-as-upheaval-hits-industry

http://www.cnc-communications.com/blog/end-of-japanese-keiretsu-but-not-yet-global-suppliers/

END OF JAPANESE KEIRETSU – ONE MORE STEP TAKEN

November 27, 2016 English By Jochen Legewie

At yesterday’s show of Nikkei FT Saturday 9 I was asked to comment on Nissan’s recent sale of its parts maker Calsonic Kansei to KKR for 4.5 bn US dollar. The deal is remarkable in many aspects but I want to focus on just one here, i.e. “Does it mark an end to the infamous vertical keiretsu relations of Japanese auto OEMs to their suppliers?”.
The Calsonic sale leaves Nissan with only one supplier, in which it owns more than 20% (Nissann Shatai). Toyota by contrast still owns more than 20% in five major suppliers. But even Toyota and Honda do no longer rely on their keiretsu suppliers as they did in the past, they increasingly also buy from outside their own groups.
Last year, Toyota for example decided in a shock for its own group firm Denso to purchase an integrated sensor module from German Continental for the world-first introduction of the “Toyota Safety Sense C” system.
Japanese OEMs increasingly request top notch technologies and global scale development and production capacities. And this is where I did a bit of research for my one minute TV comment and came across a few interesting figures.
Out of the 10 top global OEM auto parts suppliers, only two stem from Japan, i.e Denso (No. 2) and Aisin Seiki (No. 7). Germany by contrast features five in the Top 5, i.e. Bosch (No. 1), Continental (No. 4) and ZF Friedrichshafen (N0. 5) according to figures by Automotive News.
More interesting than the simple ranking is the split of sales by regions. The percentage of Asia incl. Japan out of global sales by Denso and Aisin Seiki was 61% and 71% respectively.
The respective figures for Bosch, Continental and ZF Friedrichshafen for Europe, their comparable home region, were between 47% and 49%. At the same time, all three sold at least 20-31% in each North America and Asia.
In other words: Even Japan’s biggest and strongest parts suppliers are yet much less globalized than their global peers. And this is directly linked to being a keiretsu supplier. Toyota’s partial ownership of Denso and Aisin Seiki is obviously blocking these firms from fully globalizing – yet.
The same is true for most Japanese auto parts suppliers incl. the Tier 2 and Tier 3 level. It is increasingly a problem for these firms but also for Toyota and Honda.
Even Calsonic Kansei had not managed to break free from its over-reliance on Nissan as its customer although Carlos Ghosn of its parent Nissan had always demanded exactly that.
This year, the share of Nissan in all Calsconic sales had even risen from 84% to 85%. So it is no surprise that Ghosn decided to ultimately cut loose the capital relationship and gain the freedom to purchase from the best suppliers available, i.e. globally and not only within its own keiretsu group.
You can bet that we

Semicast Research@semicast_res

Aisin and Denso are both Toyota-keiretsu. In reality @seeingmachines and @SmartEyeAB are outsiders to take Toyota DMS business as Tier 2s. Will take time. Other Japanese OEMs are "lower hanging fruit".
……………….

Aisin DMS with cadaver detection. Looks like PERCLOS, but no eye-gaze tracking. DDL (Driver Drowsiness Level) is current state-of-the-art sleep monitoring.

https://www.youtube.com/watch?v=Jui-lV_l_H0