Details

Autor(en) / Beteiligte

• Kovacs, Gabor-Daniel
• Tepelea, Laviniu Flavius
• Cret, Calin

Titel

A Comparative Study of Traditional and Deep Learning Methods in Image Depth Measuring

Ist Teil von

• 2023 17th International Conference on Engineering of Modern Electric Systems (EMES), 2023, p.1-4

Ort / Verlag

IEEE

Erscheinungsjahr

2023

Quelle

IEEE Electronic Library Online

Beschreibungen/Notizen

• This study compares traditional and deep learning methods in image depth measuring. Traditional methods often rely on manually created features and mathematical models to estimate the depth of an image, while machine learning methods use convolutional neural networks to learn considerable dataset directly from the image data. The present study evaluates the performance of both methods on a dataset of images with real depth values. The results reveal that deep learning methods prevail over traditional methods in accuracy, especially when dealing with complex and uncalibrated images. In addition, we looked into the factors that affect the performance of each method and provide an overview of their strengths and weaknesses. Overall, this study contributes with a comprehensive analogy of traditional and deep learning methods in image depth measuring and emphasizes the prospects of deep learning in supporting the field.