NVIDIA's press release states that "Jetson TX1 is the first embedded computer designed to process deep neural networks -- computer software that can learn to recognize objects or interpret information." The 3.4x2inch module includes a Tegra X1 ARM Cortex-A57 processor with 256-core NVIDIA Maxwell graphics, 4GB of LPDDR4 memory, 16GB of eMMC storage, 802.11ac WiFi and Bluetooth, and Gigabit Ethernet support.
AnandTech Article: http://www.anandtech.com/show/9779/nvidia-announces-jetson-tx1-tegra-x1-module-development-kit
The Jetson TX1 Development Kit will be available for preorder starting Nov. 12 for $599 in the United States. The kit includes the Jetson TX1 module, a carrier board (pictured below), and a 5MP camera. The stand-alone module will be available in early 2016 (for $299 in bulk).
The Jetson TK1 (not TX1) was released in 2014 to encourage the development of products based on the Tegra K1 processor. However, according to AnandTech, developers were using the Jetson TK1 outright as a production board, choosing to focus on peripheral and software development instead of system hardware development. With the new TX1, all of the I/O connectivity is provided on a carrier board, enabling rapid development on the credit-card sized TX1 module. After development is finished, the TX1 module can be directly deployed in products, such as drones.
NVIDIA used a drone application to promote the Jetson TX1
https://twitter.com/NVIDIATegra/status/664238535096926208
Comments
@JB
Allow me to get back on your comment concerning the spherical capture process.
The distribution of 6 cameras on a spherical image capture model offers a lot of possibility and more specifically the possibility of having 12 pairs of stereoscopic baselines for matching, SURF extraction or other specialized stuff. So when navigating autonomously in a dense environment (like racing in the wood) it offers multiple options to evaluate acceleration, speed and position using any significant correlation within theses captured images. This can apply as well on multiple monoscopic optical flow algorithms And yes, there are as much valuable information to be process from the side looking and backward looking cameras. Directional stereoscopic process is an anthropomorphism, and a GPU don't bother with multiple dimension and processing 12 x 12 matrix as long as you can cool it down. Never wished you hand a third eye ?!?!
Best regards JB, I appreciate your writing style, and I really wish i could be as fluent as you are. English is not my primary language, but hope you can understand my point.
@Fnoop Dogg
You make my day !!
Looks like I am not alone looking for an advanced camera system based on CSI which has intrinsic capability to sync video feeds and generate true stereoscopic buffers. The question is how many experimenters are interested in this ? I am quite certain research centers and university (MIT and ETH @ Zurich certainly will, that's what theirs systems are made of), and a couple of crazy experimenters like me will be interested.
Here`s what I suggest:
@ Jurgens; Keep the project as is for the moment, and please give me a feedback on my camera and board suggestion I posted earlier,
@JB Can you have this option detailed in the survey so we can have a good feedback on the potential of such board variation?
Final though, on the flat cable: Fnoop Dog, the Ebay Cable is not a better option because clock and signalling requires balanced lines. Jurgens's B101 could be used as an extender with HDMI cabling (right Jurgen ?) , but then you end up with a pretty big cable. Micro coax, Twisted pair could be options as well but they all require some kind of converter and that is affecting the adoption rate.
@Jurgen @Patrick
With the power of this board and the ability to handle and process multiple camera inputs, this could very well be a breakthrough for SLAM and collision avoidance - it could be so useful from a safety and regulatory point of view that it could become standard on larger UAVs.
But not if it uses FFCs. Gosh, multiple FFCs in a confined space with other sensitive electronics - might as well stick an EMP generator in there for good measure :) There are hdmi converters and frc converters like:
http://www.ebay.co.uk/itm/For-Raspberry-Pi-Camera-1m-FRC-Flexible-R...
These kits are potentially better because they're a bit more flexible and can be shielded, but they're expensive and are an absolute nightmare to put together, and still require ribbons at either end.
If there's any, any, any other way to interface the CSI please consider it :)
@Jurgen
I received the camera schematics and the architecture is quite similar to the RPICam, so you can carry on the design.
@FnoopDog
Agree with the FFC but it is a compromise to get easy access to CSI2 class devices. Concerning the Broadcom magic, RPI has designed this board with the standard CSI 2Lanes interface and the magic is that they are one of the first to have successfully implemented this technology on this market. They are now numerous chips that are CSI2 compliant, and we can expect to see them more and more with ADAS.
Just for the record, flat cable is a tradeoff between building a very specialized board supporting 6 CSI2 cameras - that is my dream - and a more general purpose configuration. The next step will be to build an adaptor board - like Jurgen's B101 that will have SerDeser chips like TI or Maxim to feed CSI cameras on Power over coax. The products could look like this :
Hi cjg,
sorry, USB 3.1 is not an option, as it requires an USB 3.1 controller in the TX1. But it only supports USB 3.0. At least 3.1 is not mentioned in the TX1 Technical Reference Manual.
Hi Fnoop Dogg,
you raise some very good points. But what is the alternative to the ribbon (FFC) cable? Unfortunately there is no standard CSI-2 connector. The advantage is that these FFC cables are fairly inexpensive. I could make them available for EUR 3 for a set of 10 cables. So if they break just replace them. I agree, EMI is an issue. And the software driver.
I guess the best solution is to use the 3Gbit serial connection via coax or STP. But this would require a custom interface board on the imager side.I think this area will require quite a bit of engineering and try and error before the perfect solution will be found.
At this point the 2 CSI-2 connectors are just an add-on for people, who wish to play with the CSI-2 bus. I guess the main interfaces are going to be the 2 USB 3.0 connectors and the 2 HDMI inputs.
Regards, Jurgen
All very exciting stuff. From a user/enthuiasist point of view (ie. I know little about the intracacies of all this stuff), please consider using anything at all other than the csi ribbon connectors, and for that matter the rpi camera. The ribbons are very fragile, prone to damage, the ends fray very very easily, they are wide and not flexible so difficult and unsightly to route around a UAV, and worst of all are unshielded and generate considerable EMI. They are the worst type of connector/cable that I have come across since getting into this uav/electronics malarky.
Also, isn't the actual processing of the rpi sensor data all done by proprietary magic inside the broadcom chip/firmware, and that bit of the firmware that hasn't been open sourced? So this would have to be reverse engineered and replicated on this board for them to be of any use. And I seem to remember someone saying that the sensor has been EOLd by the OEM, so presumably it will be replaced at some point in the not too distant future.
Perhaps you could hook up with this guy:
http://diydrones.com/profiles/blogs/360-degree-stereo-vision-and-ob...
:)
We also could start to use USB 3.1 Typ C connectors. They are small and light (but maybe mechanical not secure enough?) ...
Hi JB,
thank you for this offer. Any help is appreciated. So please go ahead and put a form together.
Regards, Jurgen
Jurgen thanks for the files I'll have a look at them this evening.
Also I already started putting together some questions for the survey and can continue with the google form if you want me to do it for you. We can then add/change/delete as required to cover all the questions we need answered for design/marketing, if you want to do it like that. BTW Just to be clear I'm not expecting anything in return other than seeing the board developed so I can buy some! ;-)
Request granted, you can proceed,many thanks Dustin
-
14
-
15
-
16
-
17
-
18
of 28 Next