CVLGIVMLOct 16, 2019

Generative Modeling for Small-Data Object Detection

arXiv:1910.07169v172 citations
Originality Highly original
AI Analysis

This addresses the problem of data scarcity in object detection for domains like medical imaging, offering a novel solution with significant performance gains.

The paper tackles object detection with limited annotated data by developing a generative model that jointly optimizes image generation and detector performance, achieving a 20% relative improvement in average precision and 50% in localization accuracy on medical and pedestrian datasets.

This paper explores object detection in the small data regime, where only a limited number of annotated bounding boxes are available due to data rarity and annotation expense. This is a common challenge today with machine learning being applied to many new tasks where obtaining training data is more challenging, e.g. in medical images with rare diseases that doctors sometimes only see once in their life-time. In this work we explore this problem from a generative modeling perspective by learning to generate new images with associated bounding boxes, and using these for training an object detector. We show that simply training previously proposed generative models does not yield satisfactory performance due to them optimizing for image realism rather than object detection accuracy. To this end we develop a new model with a novel unrolling mechanism that jointly optimizes the generative model and a detector such that the generated images improve the performance of the detector. We show this method outperforms the state of the art on two challenging datasets, disease detection and small data pedestrian detection, improving the average precision on NIH Chest X-ray by a relative 20% and localization accuracy by a relative 50%.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes