Video of type 8UC1 is not working about darknet

[prabindh]

Did you try this suggestion - "A simple way could be, to convert C1 to BGRA using something like, cvtColor(src,dst,CV_GRAY2RGB);, before feeding to the detect function."

[prabindh]

Have you been able to train Yolov3 using 1-channel, to generate a working model ?

[prabindh]

A simple way could be, to convert C1 to BGRA using something like, cvtColor(src,dst,CV_GRAY2RGB);, before feeding to the detect function.

[prabindh]

To do it efficiently with 1 channel itself without conversion, can you provide a trained model, so I can verify the detection changes ?

[shaikhibrahim951]

I am training data using Yolov2a and image type is 1 channel.
The code is giving problem in following line.
Image is getting converted to floatMat, then following process is happening.

`
if (floatMat.channels() != 3)
{
EPRINTF("Detect: channels = %d \n", floatMat.channels());
return false;
}
// Get the image to suit darknet
cv::Mat floatMatChannels[3];
cv::split(floatMat, floatMatChannels);
vconcat(floatMatChannels[0], floatMatChannels[1], floatMat);
vconcat(floatMat, floatMatChannels[2], floatMat);

__Detect((float*)floatMat.data, thresh, hier_thresh, objectCount);`

It is going to this if condition if (floatMat.channels() != 3)

Image is not passing to __detect function.

Thanks

[prabindh]

Could you also specify which camera you are using so I could try some sample data ?

[shaikhibrahim951]

cvtColor(src,dst,CV_GRAY2RGB); worked for me and able to detect desired object in frame. I am using Melaxis ToF camera.

[prabindh]

Very good and thanks for confirming. If permitted by your project, also please share some sample train images you are using.

[prabindh]

Closing as the issue is resolved.