Inside-Outside Net: Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks

TLDR

It is well known that contextual and multi‑scale representations are crucial for accurate visual recognition. The paper introduces Inside‑Outside Net (ION), an object detector that exploits information both inside and outside the region of interest. ION incorporates outside context via spatial recurrent neural networks and captures inside multi‑scale features with skip pooling, with extensive experiments guiding design choices. ION raises PASCAL VOC 2012 mAP from 73.9 % to 77.9 % and MS COCO mAP from 19.7 % to 33.1 %, winning Best Student Entry and placing third in the 2015 COCO Challenge, demonstrating that context and multi‑scale features improve small‑object detection.

Abstract

It is well known that contextual and multi-scale representations are important for accurate visual recognition. In this paper we present the Inside-Outside Net (ION), an object detector that exploits information both inside and outside the region of interest. Contextual information outside the region of interest is integrated using spatial recurrent neural networks. Inside, we use skip pooling to extract information at multiple scales and levels of abstraction. Through extensive experiments we evaluate the design space and provide readers with an overview of what tricks of the trade are important. ION improves state-of-the-art on PASCAL VOC 2012 object detection from 73.9% to 77.9% mAP. On the new and more challenging MS COCO dataset, we improve state-of-the-art from 19.7% to 33.1% mAP. In the 2015 MS COCO Detection Challenge, our ION model won "Best Student Entry" and finished 3rd place overall. As intuition suggests, our detection results provide strong evidence that context and multi-scale representations improve small object detection.