Tools
Tools
John, Calvin, Chow, Sean and Ma, YingLiang 
ORCID: https://orcid.org/0000-0001-5770-5843
(2026)
AgriFill: A Lightweight GAN-Based Image Inpainting Framework for Dynamic Object Removal in Agricultural Environments.
In: International Conference on Computer Vision and Data Mining (ICCVDM) 2026, 2026-08-15 - 2026-08-17.
Abstract
Dynamic objects such as livestock frequently occlude agricultural imagery, degrading downstream tasks including localisation, image retrieval and crop monitoring. This paper presents AgriFill, a lightweight conditional generative adversarial network (cGAN) for dynamic object removal in agricultural environments. The framework combines a U-Net generator, PatchGAN discriminator and Segment Anything Model 3 (SAM 3) masks with a weighted multi-objective loss comprising adversarial, reconstruction, masked-region and SIFT descriptor losses. Trained on synthetic agricultural imagery, AgriFill is evaluated against DeepFill v2, LaMa, Telea, Navier–Stokes, BrushNet and Stable Diffusion Inpainting. Results show the lowest masked reconstruction error, the fastest inference time and encouraging generalisation to real-world agricultural imagery, demonstrating an effective balance between reconstruction quality and computational efficiency.
| Item Type: | Conference or Workshop Item (Paper) |
|---|---|
| Faculty \ School: | Faculty of Science > School of Computing Sciences Faculty of Science |
| UEA Research Groups: | Faculty of Science > Research Groups > Data Science and AI Faculty of Science > Research Groups > Health Computing Faculty of Science > Research Groups > Norwich Epidemiology Centre Faculty of Medicine and Health Sciences > Research Groups > Norwich Epidemiology Centre Faculty of Science > Research Groups > Colour and Imaging Lab |
| Related URLs: | |
| Depositing User: | LivePure Connector |
| Date Deposited: | 06 Nov 2026 14:11 |
| Last Modified: | 06 Nov 2026 14:11 |
| URI: | https://ueaeprints.uea.ac.uk/id/eprint/105003 |
| DOI: |
Actions (login required)
 |
View Item |