Closed bglgwyng closed 11 months ago
same issue
Same issue :/
Same issue! Please @mrousavy help us.
What happens if you pass pixelFormat="rgb"
or pixelFormat="yuv"
to the <Camera>
component? Does that work?
Hello,
‘yuv’ crash my app (some internal function… it seems when VideoPipeline.kt tries to transform surface to image)
‘rgb’ works for me but there is no image bytes to copy behind… I need this info in order to implement some image post-processing with opencv…
hello everyone, with this PR there are no more YUV worries,
Basically, the problem is that the OpenGL pipeline was RGB-only.
so this pipeline was abandoned and switched to ImageWriter.
https://github.com/mrousavy/react-native-vision-camera/pull/1789
Hey- yea this PR fixes the frame processor issue once and for all, but it breaks for video recordings. Not sure why, as the ImageWriter
should be writing Frames perfectly fine to the MediaRecorder
's surface- I need to investigate this; https://github.com/mrousavy/react-native-vision-camera/pull/1799
I tested this change(#1799)
I am using react vision camera version 3.0.0
Pixel format set as "yuv" for frameprocessor
pixelFormat="yuv"
and I get this error now
com.facebook.jni.CppException: Compiling JS failed: 1:1:invalid empty parentheses '( )' Buffer size 3 starts with: 280a29 and has protection mode(s): rw-p at com.mrousavy.camera.frameprocessor.FrameProcessor.call(Native Method) at com.mrousavy.camera.core.VideoPipeline.onImageAvailable(VideoPipeline.kt:79)
It is failing in the VideoPipeline.kt
on fp.call(frame)
Hey, I have the same issue, have you guys figured out how to solve that?
Same issue @mrousavy.
fixed with 3.2.0 :)
@mrousavy Is the format supposed to be something else now? In my frame processor using 3.2.2
I still get 34
eg. ImageFormat.PRIVATE
.
edit: passing pixelFormat="yuv"
does change the format.
@mgcrea yup you need to pass the pixel format, I could maybe explain this a bit better in the docs?
Question
I found the code below from MLKit's face detector failed with the following error message.
So I investigated the image format of the
frame.image
and it wasPRIVATE
(34). I've already observed that MLKit's face detector worked fine with RNVC V2. So I suppose that this is a change from RNVC v3.I don't fully understand the description of
PRIVATE
format in the android official document linked, but it doesn't appear to be a commonly employed value. Is it a bug? Or did I miss something?I tested it on a Galaxy S20.
What I tried
No response
VisionCamera Version
3.0.0
Additional information