Robust and Explainable Fine-Grained Visual Classification with Transfer Learning: A Dual-Carriageway Framework

TL;DR

This paper introduces the Dual-Carriageway Framework (DCF), an automatic method for selecting optimal training strategies in fine-grained visual classification, providing explainability and improving accuracy through a dual-direction search and analysis.

Contribution

The paper presents DCF, a novel framework that systematically determines the best training approach and offers explainability, addressing a key gap in fine-grained visual classification.

Findings

Fine-tuning outperforms training from scratch by up to 2.13% accuracy.

DCF identifies reflection padding as the best method, increasing accuracy by 3.72%.

The framework provides quantitative and visual explanations for training strategy choices.

Abstract

In the realm of practical fine-grained visual classification applications rooted in deep learning, a common scenario involves training a model using a pre-existing dataset. Subsequently, a new dataset becomes available, prompting the desire to make a pivotal decision for achieving enhanced and leveraged inference performance on both sides: Should one opt to train datasets from scratch or fine-tune the model trained on the initial dataset using the newly released dataset? The existing literature reveals a lack of methods to systematically determine the optimal training strategy, necessitating explainability. To this end, we present an automatic best-suit training solution searching framework, the Dual-Carriageway Framework (DCF), to fill this gap. DCF benefits from the design of a dual-direction search (starting from the pre-existing or the newly released dataset) where five different training settings are enforced. In addition, DCF is not only capable of figuring out the optimal training strategy with the capability of avoiding overfitting but also yields built-in quantitative and visual explanations derived from the actual input and weights of the trained model. We validated DCF's effectiveness through experiments with three convolutional neural networks (ResNet18, ResNet34 and Inception-v3) on two temporally continued commercial product datasets. Results showed fine-tuning pathways outperformed training-from-scratch ones by up to 2.13% and 1.23% on the pre-existing and new datasets, respectively, in terms of mean accuracy. Furthermore, DCF identified reflection padding as the superior padding method, enhancing testing accuracy by 3.72% on average. This framework stands out for its potential to guide the development of robust and explainable AI solutions in fine-grained visual classification tasks.