NymPy interoperability is well supported, and with EMAN3 modules, JAX as well.

Note that converting an EMData object to a NumPy array produces an object which shares the memory of the EMData array, but when converting NumPy to EMAN2 a copy is made :

from EMAN2 import *
import numpy as np

emd=test_image()  # an EMData object
npd=to_numpy(emd)

npd*=2    # This will modify the values in BOTH the EMData object and the NumPy object!
print(emd["mean"],npd.mean())


npd2=np.zeros((128,128))
emd2=from_numpy(npd2)      # data is copied to make a new EMData object

npd2+=1   # Modifies the NumPy array, NOT the EMData object
print(emd2["mean"],npd2.mean())