How to store array of numpy array in csv file?-CodePudding

I have an image name, image path and it's corresponding cnn feature vector (a numpy array). And there are many images, meaning I have a list of names, paths, vectors (each being numpy array of dimension (1, 2048, 14, 14) and float32 as data type). I have to store and access in an efficient way. I was thinking of going with storing it in a csv file. When I did that, I am unable to convert back the numpy array from str. Kindly let me know how to proceed with this, thank you.

data = pd.DataFrame(columns=['name', 'path', 'vector'])
data['name'] = image_names
data['path'] = image_paths
for i in range(len(image_paths)):
   out = encoder(image_paths[i] #(1,2048,14,14)
   data['vector'][i] = out
data.to_csv('encoding.csv', sep = ',', na_rep='Unknown')

dr =pd.read_csv('encoding.csv')
for i in range(dr.shape[0]):
    c = dr['vector'][i]

c is a str object. Unsure how to convert this to numpy array.

CodePudding user response：

You could save the numpy array using numpy.save as shown in the docs here. You could save the image name and image path as a .json file (and just make the name of the .npy file the same as the name as the .json file to allow for easier searching).