Hi,
I currently have the dm365 EVM with the face detection hardware and I am running the demo program that came with the DVSDK 2.10. This demo program takes a file as input and runs it through the face detection hardware and prints the output. I have been able to successfully run this program on many different images (a face detection database).
The next step was to start to run the face detection algorithm on incoming video, so I started with the capture on the fly yuv linux driver example included in the dvsdk. I set up the resizers to output a NV12 colorspace 320x240 image and passed this image to the driver. The detection only works if I call the FACE_DETECT_EXECUTE ioctl() twice for the same image. For example, I set up the hardware params and buffers and then call execute. The first call to execute returns no faces but when I call execute immediately after this first execute I get the correct result.
Has anyone else observed this behavior?
Thanks,
Randy Scheifele