Image compression approach for improving deep learning applications

<span lang="EN-US">In deep learning, dataset plays a main role in training and getting accurate results of detection recognition objects an image. Any model needs large size to be more accurate, where improving the is one most research problems that enhancement. In this paper, image compression approach was developed reduce improve classification accuracy for trained using convolutional neural network (CNN), speeds up machine learning process, while maintaining quality. The revealed best scenario models provided good acceptable had following parameters: 80×80 size, 10 epochs, 64 batch 40 images quality (images compressed 60%), gray mode. For Dog vs Cat used, time 48 minutes, 86%, 317 MB on storage device. This makes 58% original image’s dataset, saves 42% space reduces processing resources consumption.</span>