DCDS-Net: Deep transfer network based on depth-wise separable convolution with residual connection for diagnosing gastrointestinal diseases

Sohaib Asif, Ming Zhao, Fengxiao Tang, Yusen Zhu

Abstract: Highlights•We propose DCDS-Net based on depth-wise separable convolutional neural network.•Data augmentation is used to improve performance and reduce overfitting.•We have analyzed the effect and significance of block wise fine-tuning strategy.•Extensive experiments and detailed comparative analysis were presented.•The proposed model shows excellent classification performance even on small dataset.