Summary of Leveraging Latent Diffusion Models For Training-free In-distribution Data Augmentation For Surface Defect Detection, by Federico Girella et al.
Leveraging Latent Diffusion Models for Training-Free In-Distribution Data Augmentation for Surface Defect Detection
by Federico Girella, Ziyue Liu, Franco Fummi, Francesco Setti, Marco Cristani, Luigi Capogrosso
First submitted to arxiv on: 4 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel approach to data augmentation for defect detection, introducing DIAG (Diffusion-based In-distribution Anomaly Generation) pipeline. Unlike conventional image generation techniques, DIAG incorporates human-in-the-loop feedback through text descriptions and region localization of possible anomalies. This strategic shift enhances interpretability and fosters a robust feedback loop. The model operates in zero-shot manner, achieving superior performance without fine-tuning. Evaluation on the challenging KSDD2 dataset shows an improvement in AP of approximately 18% when positive samples are available and 28% when they are missing. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary DIAG helps machines learn what is not a normal sample but can accurately identify what a defect looks like. This paper introduces a new way to make machines better at finding defects by adding fake data that experts help create. The machine doesn’t need training, just expert input on what might be wrong with the images. This makes it faster and more accurate. |
Keywords
» Artificial intelligence » Data augmentation » Diffusion » Fine tuning » Image generation » Zero shot
