You should also know that 4K resolution is not a wise choice for the Nano because the Nano will be to weak to make serious treatments with such resolution sensor. FullHD is more suitable for the Nano.
I strongly disagree. Even back in 2019 there were projects where Jetson Nano was useful to process 4K resolution video, for example: FPV in 4K | befinitiv . These days Jetson Nano 2G is even better choice for relatively simple projects which need high resolution but do not need 4G RAM.
Of course, 4K can be used with AI too, for example, process low resolution frames at relatively high FPS, and when necessary to get more details, crop area(s) of interest from 4K (for example, if a movement was identified in low resolution stream, but the object is too small to recognize, then using its bounding box, take a crop from 4K frame to get more details). Or process 4K frames at low FPS - if it is sufficient, then Jetson Nano is very good choice considering its relatively low price. There are many other possibilities as well.
It is of course obvious that Jetson Nano cannot process 4K@90FPS even with overclocking. But not every project needs to process full 4K frames at high FPS. There is definitely a need for high resolution sensors capable of 30 or more FPS for Jetson Nano. If you personally have no use of high resolution cameras in your Jetson Nano projects, just buy a camera module you need, but please do not claim that others do not need high resolution sensors.
IMX385 is a very good sensor under low light conditions. Many astronomers have tested it and this sensor is a must have, even if it is FSI sensor. STARVIS is just a marketing name and you should try to study sensor specifications instead of commercial argument.
This is true. Even though IMX385 is not STARVIS sensor, its light sensitivity is better than most STARVIS sensors. Here is a useful table where SNR1s values are shown for many Sony sensors (the website is non-English, but the table itself should be clear to anyone who can understand numbers). Also, the table contains sensor size and year of release. For example, IMX415 (1/2.8") has SNR1s 0.79@4K and IMX385 (1/2") has SNR1s 0.13@FullHD; IMX415 with digital 2x binning (in other words, FullHD resolution) will have SNR1s 0.395. In other words, IMX385 has 3 times better light sensitivity than IMX415 at the same resolution.
But this is not as simple as that. IMX415 may use smaller and lighter lenses, and with sufficient lighting will provide 4 times better resolution. So I do not think it is fair to compare IMX385 and IMX415. They are too different. Another example: the most sensitive FullHD sensor is IMX482 (0.07 SNR1s), but it would not be fair to compare it with IMX385 for the same reason - IMX385 can use smaller and lighter lenses, and it is cheaper. Example of fair comparison, is to compare IMX290 and IMX327 - they have the same size and the same resolution, but IMX327 has 1.28x better light sensitivity. But many applications do not need the best light sensitivity, so this is why IMX307 exists, which is slightly worse in therms of light sensitivity than deprecated IMX290 but it is cheaper and newer.
If choosing a particular sensor is an easy task for you because you know exactly what you want for your usecase(s), this is great. But please keep in mind that others may have different priorities and needs for their projects.