OpenIPC / majestic

Majestic Community edition integration kit
MIT License
32 stars 5 forks source link

WYZE v3 (T31) exposure problems in the daytime #102

Open mariarti opened 1 year ago

mariarti commented 1 year ago

Required information

Issue description

Exposure problems in the daytime. At night / cloudy time the picture is normal.

Expectations

Good picture at any time of day (any lighting scene)

Steps to reproduce

  1. Just turn on

Information to attach

openipc-2 3 02 11-lite-t31zx-gc2053-1676393102-snapshot4cron openipc-2 3 02 11-lite-t31zx-gc2053-1676392202-snapshot4cron openipc-2 3 02 11-lite-t31zx-gc2053-1676389501-snapshot4cron

Assumption https://t.me/c/1223369578/13838

flyrouter commented 1 year ago

@dimerr Please see this problem, what we can fix here. This is our chance to run a popular device, very many hits.

dimerr commented 1 year ago

take gc2053.bin from stock fw and try

ialexlog commented 1 year ago

Replacing gc2053.bin does not give the desired result. The only difference is the absence of the pink ring in the middle of the snapshot, and that's it.

As it was found, this particular kind of highlighting is caused by the snapshot script. If a snapshot is taken externally (e.g., curl, wget), the problem does not manifest itself. The assumption was made that there is not enough memory when the script is called. This is most likely the case. At the moment a workaround in the form of an external snapshot rather than directly from the camera is used. I would be happy for it to work normally without any workarounds.

gtxaspec commented 10 months ago

This PR https://github.com/OpenIPC/firmware/pull/966 has fixed exposure issues for me. Maybe you can try the latest master and see if you still have issues.

damienwolf07 commented 3 months ago

Replacing gc2053.bin does not give the desired result. The only difference is the absence of the pink ring in the middle of the snapshot, and that's it.

As it was found, this particular kind of highlighting is caused by the snapshot script. If a snapshot is taken externally (e.g., curl, wget), the problem does not manifest itself. The assumption was made that there is not enough memory when the script is called. This is most likely the case. At the moment a workaround in the form of an external snapshot rather than directly from the camera is used. I would be happy for it to work normally without any workarounds.

Ah, For the snapshot script, I have found out that its when we pull the sensor model, Is when the flash occurs. I would guess when we pull the sensor model (the command that returns 'gc2053') the sensor is temporarily stopped so we can get it's model, Thus ruining the auto exposure loop. A workaround for me was making the sensor model variable static.

EDIT: This command is run before we take a snapshot, So it takes a snapshot while the exposure goes back to normal.