Image de-fencing using cGANs

- Divyanshu Gupta, Shorya Jain, Utkarsh Tripathi, Pratik Chattopadhyay, Lipo Wang

Abstract -

Image de-fencing is one of the important aspects of recreational photography in which the objective is to remove the fence texture present in an image and generate an aesthetically pleasing version of the same image without the fence texture. In this paper, we develop an automated and effective technique for fence removal and image reconstruction using conditional Generative Adversarial Networks (cGANs). These networks have been successfully applied in several other domains of Computer Vision focusing on image generation and rendering. Our approach is based on a two-stage architecture involving two cGANs in succession, in which the first cGAN generates the fence mask from an input fenced image, and the next one generates the final de-fenced image from the given input and the corresponding fence mask obtained from the previous cGAN. Training these networks is carried out independently using suitable loss functions, and during the deployment phase the above two networks are stacked together in an end-to-end manner to generate the de-fenced image from an unknown test image. Extensive qualitative and quantitative evaluation on challenging data sets emphasize the effectiveness of our approach over state-of-the-art de-fencing techniques. The data sets used in the experiments have also been made available for further comparison.

Download-

1. Dataset (Full dataset has been uploaded now...!!!)

2. Pre-trained generator weights (A research paper has been communicated on this work and pre-trained models will be made available once paper gets accepted.)

Qualitative Results-

Fig. Qualitative results for our image de-fencing network. (a) Original fenced image, (b) Generated mask by fence mask generator, and (c) Final de-fenced image.