BPI-M3 compile Opencv 3.1.0 source code

Hey @sinovoip: You wrote again so much I don’t understand but the intructions above are still wrong. Why don’t you correct them? Don’t you fear users could try them out and fail? Do you care about your users? And their time? Or are you convinced that no one will ever try out the stuff you waste your time with like above?

1 Like

where is wrong? did you test it?

Read above? One time emmc/ is too much, the other time a # is missing. Did you test it?

you can test this image at frist.


hehe .this is from my shell .in test time ,#and \ do not need. emmc is my dir .

1 Like

C’mon: If you’re at ~/emmc and then do a

sudo git clone https://github.com/Itseez/opencv.git

Then now opencv is in


If you now do the following while still being in ~/emmc

sudo mkdir emmc/opencv/build
cd emmc/opencv/build

then opencv is still in ~/emmc/opencv but you created and changed into ~/emmc/emmc/opencv/build while you wanted to be at ~/emmc/opencv/build instead. There’s is one step missing (cd ..) or the occurences of emmc/ are too much.

I wrote exactly this two times already but it’s obviously absolutely useless since you don’t seem to take care of the difference between ~/emmc/emmc/opencv/build (wrong) and ~/emmc/opencv/build (right). The ‘instructions’ above simply can not work and that’s just one of the mistakes. And it’s always the same: As soon as @sinovoip posts instructions they’re partially wrong so exactly nobody can benefit from. So why do you waste your time providing useless instructions again and again?

Same here: [BPI-M3] how to use fatload uEnv.txt & script.bin & kernel uImage Still wrong: scrip=script.bin can NOT work! You’ve been told several times and you simply don’t give a sh*t. There’s still a t missing. You not even fix such simple mistakes. And this pretty much describes the entirety of the catastrophic state of documentation and the lousy support situation.

CN6 / CSI connector not populated so you’re showing pictures made with an USB cam there as proof that BPi M3 works with a dual camera setup. Always fakes… :frowning:

I’ve never seen the proof that any of your boards really works with a CSI camera :slight_smile:

Make process gave error under ubuntu… :frowning:

this is for rasbian,Ubuntu and debian have many error

tkaiser @
at sudo mkdir emmc/opencv/build means emmc/opencv mkdir build

you can test it

if you have a time can you try to install ros with this link? because I sholud use opencv with ROS. My internet download speed is very slow I cant download rasbian now.

If install successful I will download rasbian it this night.

sorry my nearest work is opencv and uav .you can do it yourself

Hmm… thanks for your helps to contribute this topic. Lastly, I want to ask did you used opencl with your application. If yes, which application did you use and how many fps did you take? If you can answer I will be very happy … :slight_smile:

are you test this Ros image ???

i have test this image but opencv unsucseece

@sinovoip I cannot test it because I don’ t have a download link. I m waiting Google Drive link.

@zhiqing_yu I tried rasbian image opencv succesfully installed just. But opencl device cannot find. I am trying to use optical flow lukas kanade algorithm. But banana pi is working 1 or 2 fps.

bpi m3 use Allwinner A83t SGX544 MP2 have 44.8 gflops

,opencl need trenclate cpu ram gpu ,at the time opencl have much eer .i sugest you use cpu only down pix up fps

This is not true. A31/A31s used SGX544MP2, A83T contains only SGX544MP1 (half that fast) and you can’t use OpenCL ‘out of the box’ anyway. Just read through the forums of responsible vendors that used these GPUs in their SoCs to get the idea (TI or NXP for example).

Everthing that’s written in datasheets or marketing material regarding BPi M3 is either irrelevant or wrong :slight_smile:


So you believe every BS you find somewhere on the web? GPU is wrong there, MHz are wrong, GFLOPS are wrong.

Look into the datasheet http://linux-sunxi.org/A83T#Documentation and look for Antutu 3D scores of A83T equipped devices. GPU clockspeed will be reduced brutally since A83T is always overheating:

Did you ever tried this out? Just do some GPU intensive stuff and watch dmesg output: GPU clock will be reduced even with mild workloads.