# Improved Inference via Deep Input Transfer

## TL;DR

This paper introduces a novel input modification technique for image segmentation that enhances accuracy by leveraging network gradients, achieving notable improvements across multiple medical datasets without altering the network architecture.

## Contribution

The paper presents a new paradigm that boosts segmentation performance by modifying inputs using gradient information, avoiding changes to the network structure.

## Key findings

- Achieved 5.8% improvement on ISIC 2017 dataset
- Achieved 0.5% improvement on Shenzhen Chest X-Ray dataset
- Achieved 4.8% improvement on CVC-ColonDB dataset

## Abstract

Although numerous improvements have been made in the field of image segmentation using convolutional neural networks, the majority of these improvements rely on training with larger datasets, model architecture modifications, novel loss functions, and better optimizers. In this paper, we propose a new segmentation performance boosting paradigm that relies on optimally modifying the network's input instead of the network itself. In particular, we leverage the gradients of a trained segmentation network with respect to the input to transfer it to a space where the segmentation accuracy improves. We test the proposed method on three publicly available medical image segmentation datasets: the ISIC 2017 Skin Lesion Segmentation dataset, the Shenzhen Chest X-Ray dataset, and the CVC-ColonDB dataset, for which our method achieves improvements of 5.8%, 0.5%, and 4.8% in the average Dice scores, respectively.