A139 - 3 channel dashcam - early impressions/review

Thanks. Not that it matters as Novatek don't seem to like to make their spec sheets publicly available, or am I just looking at the wrong places?
Even when you do have the confidential Novatek spec. sheets, it is hard to see much of interest in them! What it is capable of is mainly a matter of processing power, how many channels you want and what resolutions you want to use.
 
Even when you do have the confidential Novatek spec. sheets, it is hard to see much of interest in them! What it is capable of is mainly a matter of processing power, how many channels you want and what resolutions you want to use.

Well it's interesting to see how many hardware encoders and decoders are in there, and how the streams can be configured. HiSilicon publishes some information on that, even if it's largely meaningless because they're not as powerful. Novatek units seems to be much more powerful. Knowledge of power dissipation and manufacturing process is also a novelty.
 
@jokiin Yeah that's not very helpful at all, unless there's more to see after that. HiSilicon's Hi3559 V200 found in the DDPai Mini 5 for example mentions the below:

Video Encoding and Decoding Performance
  • Maximum resolution for H.265/H.264 encoding and decoding: 3840 x 2160
  • H.265/H.264 encoding and decoding performance:
    • 3840 x 2160@30 fps + 720p@30 fps encoding​
    • 3840 x 2160@30 fps decoding​
  • Maximum resolution for JPEG encoding and decoding: 8192 x 8192
  • Maximum JPEG encoding and decoding performance: 16 MP (4608 x 3456)@10 fps
  • Multiple bit rate control modes such as CBR, VBR, and FIXQP
  • Maximum bit rate for H.265/H.264 encoding output: 100 Mbit/s
  • Encoding of eight ROIs

From here we know if the files produced are much larger than 12.5 MB/sec, it likely has been padded by the CPU and will not contain very much more visual information than if there was another setting that produced a file which is around 12.5 MB/sec. Since the CPU is 2x Cortex A7 @ 900 MHz, the CPU itself is unlikely able to assist very much in this task. The datasheet also mentions a couple of hardware processing capabilities under the ISP section with stuff like shake compensation, noise reduction, methods of WDR supported, etc.

But perhaps the most interesting is this. Opens up a large field of possibilities for the dashcam developer depending on how hard they want to work their software development team. Maybe a future 70Mai model might be able to highlight some of these capabilities.

Intelligent Analysis
  • Facial recognition, target detection and tracking, and scenario identification
  • Neural network acceleration engine with processing performance up to 0.4 TOPS
 
Back
Top