Siudya / ORB_FPGA

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Change input image size

chi-an1997 opened this issue · comments

Thanks for your wonderful work. I am trying to change input image size from (640,480) to (2560,2560). I change WIDTH and HEIGHT in xf_config_params.h. I also modify src_buf in python as well. Is there any place I ignore or is there any suggestion for me? Thanks a lot.

You need to alter WIDTH and HEIGHT in headers of every single hls IP. Specifically, they are HLS_myGaussianFilter/xf_gaussian_filter_config.h, comFAST/xf_config_params.h, and comDescriptor/comDescriptor.h.
By the way, for 2560×2560 may need much more resources, you should choose a device better than xc7z020clg400.

Thanks for your reply. I can run 2560×2560 on pynq-z2. In opencv, FAST algorithm has two parameters( nfeatures and fastThreshold ). Can I change threshold in your code? If it is possible, where should I modify?
By the way, in jupyter notebook, "print(des_buf[97528][0:7])" can only generate 7 int32 numbers instead of 8 int32 numbers.

You can changed the parameter _threshold when calling xFfast7x7(line 878 in myFAST.cpp). I don't get what "nfeatures" is stand for. This FAST detector is nearly the same as the one in opencv with cv.FAST_FEATURE_DETECTOR_TYPE_9_16. And thanks for the correction on the range, it was my negligence.