search

xxlong0 / NeTO

Neural reconstruction of transparent objects
28 stars 1 forks source link

how can i get “screen_point.npy” file？ #2

Open qiuyuenheng opened 10 months ago

qiuyuenheng commented 10 months ago

Hello author, I am interested in your research direction and would like to compare your code results with my experiment, but I am not sure how to use your code to train my dataset. I saw that at the end of the "datao. py" file, it seemed to generate a "screen_point. npy" file, but when I set my own file path, I found:

ret, corners = cv2.findChessboardCorners(gray, (w, h), flags=cv2.CALIB_CB_ADAPTIVE_THRESH + cv2.CALIB_CB_FAST_CHECK + cv2.CALIB_CB_NORMALIZE_IMAGE)

This command line did not output the correct result. Can you provide specific instructions on how to operate?

LRussianStand commented 10 months ago

If you calculate screen_point.npy from captured images, you need to run the process_transparent.py file. If you get screen_point.npy from the .h5 file, run dataoP.py. The inability to detect corner points may be related to the image clarity and the accuracy of the checkerboard. This paper uses purchased checkerboards.