Abstract:In the current study, the capability of pre-trained Deep Convolutional Neural Network (DCNN) by ImageNet features is proposed for categorization of brain tumors by utilizing MR images. The pre-trained models like ResNet50, InceptionV3, Xception, DenseNet121, MobileNetV3Large, EffcientNetB0, EfficientNetV2L, EfficientNetV2B0 have been exploited for classification purpose. The selection criteria is based upon the diverse proficiency each model depicts. For e.g., EfficientNetB0, ResNet50, MobileNet, Xception employ lower number of training parameters that makes them time efficient whereas, VGG16 though has higher number of training parameters, while increasing the training time without compromising with accuracy. The AlexNet on the other hand also has a reduced number of parameters in comparison to Google's Inception module which is more memory efficient. AlexNet takes more memory for training. The EfficientNet's are based on inverted residual blocks of MobileNetV2 with addition to Squeeze and Excitation blocks which makes them highly efficient for feature extraction. Therefore, in this research study, the above stated pre-trained models are used in by retaining the ReLU as an activation function due to its unbounded nature which also helps the model from Vanishing Gradient problem. By hyper tuning the top layers of different pre-trained models by adding Global Average Pooling layer and Dropout layer along with Fully Connected layer with classifier as SoftMax increases the overall efficiency and also reduces overfitting. The proposed comparative study shows the working of different pre-trained DCNN models for classifying brain tumors. The proposed experimental study is performed on three different databases i.e., Kaggle, BraTS 2018, and Real time dataset acquired from PGIMER and comparison analysis. The comparison analysis with existing methods as well as statistical analysis is also performed. It is observed from the results that EfficientNetB0 has outperformed all the existing methods. The pre-trained EfficientNetB0 architecture by hyper tuning the parameters, the testing accuracy has increased by 2.14% on Kaggle and increased the classification accuracy by 3.98% on BraTS dataset. On Real time dataset comprising of Glioblastoma Multiforme and Oligodendroglioma, highest accuracy of 97.32% is achieved among all other models.

Comparison of transfer-learning for lightweight pre-trained model on image classification

Research on Image Classification Method of Features of Combinatorial Convolution

Research on Image Classification of Lightweight Convolutional Neural Network

Comparable Study Of Convolutional Neural Networks In Classification And Feature Extraction Applications

Transfer Learning on EfficientNet for Remote Sensing Image Classification

Image recognition based on lightweight convolutional neural network: Recent advances

Comparative Analysis of Different Deep Convolutional Neural Network Architectures for Classification of Brain Tumor on Magnetic Resonance Images

A comparison between VGG16, VGG19 and ResNet50 architecture frameworks for Image Classification

A Study on CNN Transfer Learning for Image Classification

Benchmark analysis of popular ImageNet classification deep CNN architectures

Establishing a stronger baseline for lightweight contrastive models

Pay Attention to Convolution Filters: Towards Fast and Accurate Fine-Grained Transfer Learning

Performance of Transfer Learning Model vs. Traditional Neural Network in Low System Resource Environment

Image recognition in depth: comparative study of CNN and Pre-trained VGG16 architecture for classification tasks

Low-res MobileNet: An efficient lightweight network for low-resolution image classification in resource-constrained scenarios

Which Backbone to Use: A Resource-efficient Domain Specific Comparison for Computer Vision

EfficientNet-eLite: Extremely Lightweight and Efficient CNN Models for Edge Devices by Network Candidate Search

Lightweight Network Research Based on Deep Learning

Connection Reduction of DenseNet for Image Recognition

Comparison of Land Use Classification Based on Convolutional Neural Network