Is there some way of sampling a numpy array with float indices, using bilinear interpolation to get the intermediate values? For example, given the 1D array: arr=np.array([0,1]) I would like arr[0.5] to return 0.5, since that index lies between 0 and 1. For a 2D example: arr=np.array([[0,1],[2,3]]) arr[0.5, 0.5] should return 1.5. In pytorch this functionality is provided by torch.nn.grid_sample, I'd like to compare performance to doing this in numpy for my application. Continue reading...
