如何使用 h5py 构建数据集？

Question

I have the following dataset :

class category : 1,2,3 each class contain rgb images 320x240

dataset
   |---- training_set
         |---- 1
               |--- rgb_1.png
               |--- rgb_2.png
         |---- 2
               |--- rgb_1.png
               |--- rgb_2.png
         |---- 3
               |--- rgb_1.png
               |--- rgb_2.png
   |---- test_set
         |---- 1
               |--- rgb_1.png
               |--- rgb_2.png 
         |---- 2
               |--- rgb_1.png
               |--- rgb_2.png
         |---- 3
               |--- rgb_1.png
               |--- rgb_2.png

Because when I training the model In google colab it is very slow than my computer, I think because the dataset is in google drive, so I try another solution to create the dataset using h5py.

Please, How to build this dataset with h5py ?

Answer 1

The short answer is use create_dataset() . Everything you need is in the link provided by @jakub. To populate the dataset, you need to create a NumPy array from your image. I don't know the best way to do that. You will need to figure that part out. For my example, I used cv2.imread() .

You can get started with this simple piece of code:

import h5py, cv2
import numpy as np
# Create an array for image data
img_arr = cv2.imread(imgFile)
# returns a np array 
with h5py.File('imagedata.h5','w') as h5f :
    dset = h5f.create_dataset("image1", data=img_arr)