Raspberry Pi 5/HQ sensor - new stuff

Updated results from my bench:

{‘size’: (4056, 3040), ‘format’: ‘RGB888’, ‘stride’: 12224, ‘framesize’: 37160960}
{‘size’: (4056, 3040), ‘format’: ‘BGGR16_PISP_COMP1’, ‘stride’: 4096, ‘framesize’: 12451840}
Metadata only Spf: 0.09040772914886475 Fps: 11.061001193309554
Make_array: Spf: 0.10541331768035887 Fps: 9.486467383867616
OpenCV Jpeg encode : Spf: 0.17197253704071044 Fps: 5.81488194108152
DNG encode : Spf: 0.868503427505493 Fps: 1.1514059338512803
Multi DNG encode : Spf: 1.0001626968383788 Fps: 0.9998373296275764
Multi Jpeg encode : Spf: 0.1040182113647461 Fps: 9.613701167129667

{‘size’: (2028, 1520), ‘format’: ‘RGB888’, ‘stride’: 6144, ‘framesize’: 9338880}
{‘size’: (2028, 1520), ‘format’: ‘BGGR16_PISP_COMP1’, ‘stride’: 2048, ‘framesize’: 3112960}
Metadata only Spf: 0.02052018642425537 Fps: 48.7325007348849
Make_array: Spf: 0.028426265716552736 Fps: 35.17873258384747
OpenCV Jpeg encode : Spf: 0.048226809501647955 Fps: 20.7353546778961
DNG encode : Spf: 0.2283710241317749 Fps: 4.3788392323492795
Multi DNG encode : Spf: 0.2800793170928955 Fps: 3.5704171603229247
Multi Jpeg encode : Spf: 0.024844217300415042 Fps: 40.25081522625767

{‘size’: (4056, 3040), ‘format’: ‘RGB888’, ‘stride’: 12224, ‘framesize’: 37160960}
{‘size’: (4056, 3040), ‘format’: ‘SBGGR16’, ‘stride’: 8128, ‘framesize’: 24709120}
Metadata only Spf: 0.08476526737213134 Fps: 11.79728479602218
Make_array: Spf: 0.11319601535797119 Fps: 8.834233226652007
OpenCV Jpeg encode : Spf: 0.17975094318389892 Fps: 5.563253145085997
DNG encode : Spf: 0.14657435417175294 Fps: 6.822475907539867
Multi DNG encode : Spf: 0.12432038784027101 Fps: 8.043732949778256
Multi Jpeg encode : Spf: 0.1241300344467163 Fps: 8.056068013332077

Notes:

  • For the HQ camera The PI5 clearly opens up the possibility of capturing in full 4K resolution with a decent fps for those who consider it important.
  • In half resolution the fps will be even better!
  • For a DNG capture as explained in the previous post, you should absolutely not use compressed mode.
  • Controlling the pi via the network by a connected PC is perhaps less useful, the duration of transmission of the frame on the network is probably not very different from the transmission on a connected SSD? In my opinion this remains convenient for ease of use.
2 Likes