Summary of Dataset Enhancement with Instance-level Augmentations, by Orest Kupyn et al.
Dataset Enhancement with Instance-Level Augmentations
by Orest Kupyn, Christian Rupprecht
First submitted to arxiv on: 12 Jun 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 introduces a novel data augmentation technique that leverages pre-trained latent diffusion models to expand existing datasets. The method, called instance-level data augmentation, repaints parts of the image at the level of object instances using conditional diffusion models with depth and edge maps control conditioning. This approach improves the performance and generalization of state-of-the-art models for salient object detection, semantic segmentation, and object detection tasks. The method is applicable to any segmentation or detection dataset and can also be used for data anonymization by redrawing privacy-sensitive instances. The authors release fully synthetic and anonymized expansions for popular datasets like COCO, Pascal VOC, and DUTS. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine taking a picture of a dog, then changing its color to blue or replacing its tail with a new one. This paper shows how to do something similar to lots of pictures at once. They developed a way to take an existing dataset of images and add more variety by “repainting” parts of the image. This helps make computer models better at recognizing objects, people, and things in images. It’s like training a model on a new set of images without actually taking all those new pictures. The authors also show how this technique can be used to protect people’s privacy in these images. |
Keywords
» Artificial intelligence » Data augmentation » Generalization » Object detection » Semantic segmentation
