Impact of fully connected layers on performance of convolutional
neural networks for image classification
URL
Summary
This paper attempts at finding the relationship between Fully Connected (FC) layers
with some of the characteristics of the datasets. This paper tries to formalize these terms along with answering the following questions. (i) What is the impact of deeper/shallow architectures on the performance of the CNN w.r.t. FC layers?, (ii) How the deeper/wider datasets influence the performance of CNN w.r.t. FC layers?,
and (iii) Which kind of architecture (deeper/shallower) is better suitable for which kind of (deeper/wider) datasets
Key Points
Popularity of Convolutional Neural Networks (CNN) is growing significantly for various application domains related to computer vision.
Careful selection of network width (number of neurons in FC layers, number of filters in convolution layers) and network depth (number of trainable layers) plays a vital role in designing deep neural networks in order to obtain
better performance.
We performed a systematic study to observe the effect of deeper/shallower architectures on the performance of CNNs with varying number of FC layers.
We observed the effect of deeper/wider datasets on the performance of CNN w.r.t. FC layers.
We developed four CNN models (CNN-1, CNN-2, CNN-3, and CNN-4) to conduct the experiments.
Our experimental results demonstrate that the number of FC layers and the number of neurons in those layers have a significant effect on the accuracy of the model.
Title
Impact of fully connected layers on performance of convolutional neural networks for image classification
URL
Summary
This paper attempts at finding the relationship between Fully Connected (FC) layers with some of the characteristics of the datasets. This paper tries to formalize these terms along with answering the following questions. (i) What is the impact of deeper/shallow architectures on the performance of the CNN w.r.t. FC layers?, (ii) How the deeper/wider datasets influence the performance of CNN w.r.t. FC layers?, and (iii) Which kind of architecture (deeper/shallower) is better suitable for which kind of (deeper/wider) datasets
Key Points
Popularity of Convolutional Neural Networks (CNN) is growing significantly for various application domains related to computer vision. Careful selection of network width (number of neurons in FC layers, number of filters in convolution layers) and network depth (number of trainable layers) plays a vital role in designing deep neural networks in order to obtain better performance. We performed a systematic study to observe the effect of deeper/shallower architectures on the performance of CNNs with varying number of FC layers. We observed the effect of deeper/wider datasets on the performance of CNN w.r.t. FC layers. We developed four CNN models (CNN-1, CNN-2, CNN-3, and CNN-4) to conduct the experiments. Our experimental results demonstrate that the number of FC layers and the number of neurons in those layers have a significant effect on the accuracy of the model.
Citation
Repo link