Rescaling pixels positions after augmented resolution of the movie

Question

I need help understanding the rescaling of coordinates pixel position (X,Y) of the center of an object in an image after rescaling from 64x64 to 640x640. I explain: To use a deep learning tracking python code, I had to change the resolution of the film I have because the code doesn't work with such a small size, so firstly, I resized the film from 64x64x50000 to 640x640x50000 (50000 is the number of the images). After calculating the objects' positions (X,Y), I retrieved a csv file in which the positions are in the format of 640x640, but I want to have this information in the original scale (64x64).

I believe that the cross product isn't really correct, I mean doing

newX = (currentX/currentWidth)*newWidth
newY = (currentY/currentHeight)*newHeight 

It is not the correct way to downsample or have the real coordinates of the original images. When calculating the speed I found a shift and differents results comparing to another method I used.

Here and example of the CSV file I retrieved and all I want is to convert the result from the 640 x 640 resolution to the 64x64 resolution

Could anyone help me understand or have any idea how can I get the true position, please?