Abstract To significant applications in robotics, autonomous driving, visual surveillance, object recognition is a crucial study area in pattern recognition. The literature introduces various computer vision methods. There are many difficulties, such as imbalanced dataset & similar shapes of various items. Moreover, they deal with irrelevant feature extraction, which decreases classification performance and enhances calculation. We proposed a totally automatic computer image pipeline in this article. In proposed strategy, original data augmentation is done to balance the categorized objects. A Convolutional Neural Network (CNN) was afterwards taken into consideration and tweaked in accordance with the chosen dataset (Caltech101). The improved model extracts characteristics and was trained via transfer learning. A few unnecessary pieces of information were deleted from the collected characteristics using an Improved Whale Optimization Algorithm (IWOA). The total precision can be enhanced by using auto encoder-based dimensionality reduction, vector-based pixel reconstruction, and loss identification. The categorization procedure for color photos of people is implemented using CNN approach. The accuracy & effectiveness to the proposed method have enhanced according to the performance evaluation results when compared to the existing techniques.
Alan : Mühendislik
Dergi Türü : Uluslararası
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|