WSOD2: Learning Bottom-Up and Top-Down Objectness Distillation for Weakly-Supervised Object Detection

Abstract

We study on weakly-supervised object detection (WSOD) which plays a vital role in relieving human involvement from object-level annotations. Predominant works integrate region proposal mechanisms with convolutional neural networks (CNN). Although CNN is proficient in extracting discriminative local features, grand challenges still exist to measure the likelihood of a bounding box containing a complete object (i.e., "objectness"). In this paper, we propose a novel WSOD framework with Objectness Distillation (i.e., WSOD 2 ) by designing a tailored training mechanism for weakly-supervised object detection. Multiple regression targets are specifically determined by jointly considering bottom-up (BU) and top-down (TD) objectness from lowlevel measurement and CNN confidences with an adaptive linear combination. As bounding box regression can facilitate a region proposal learning to approach its regression target with high objectness during training, deep objectness representation learned from bottom-up evidences can be gradually distilled into CNN by optimization. We explore different adaptive training curves for BU/TD objectness, and show that the proposed WSOD 2 can achieve state-of-the-art results.

Reference-Based Defect Detection Network

“…C ONVOLUTIONAL neural networks (CNNs) have been widely adopted in computer vision tasks, including category classification [1]- [4], instance detection/segmentation [5]- [9], semantic segmentation [10]- [12] and cross-modality understanding [13]- [15]. The excellent performance of CNNs credits to the powerful network architecture [1]- [3], the effective mechanism design [16], [17], the development of hardware devices [18], and the large scale datasets with annotations [19], [20].…”

Efficient Weakly-Supervised Object Detection With Pseudo Annotations

“…A common procedure adpoted in WSOD methods is to train a fully-supervised obejct detector using annotations generated from the WSOD detection results. Inspired by this procedure, some works [12]- [14] try to introduce a regression branch to the MIDN directly, where the annotations are mined in various ways. In the approach proposed by Zeng et al [12], pseudo annotations are mined based on low-level image features.…”

“…Inspired by this procedure, some works [12]- [14] try to introduce a regression branch to the MIDN directly, where the annotations are mined in various ways. In the approach proposed by Zeng et al [12], pseudo annotations are mined based on low-level image features. Differently, in [13], instance-level annotations are extracted online from the results of MIDN.…”

A weakly supervised object detection approach using point annotation

“…MELM [17] 通过最小化局部、全局熵进行目标 检测. WSOD2 [19] 采用自适应线性组合, 结合自下而上的目标线索和自上而下的类别置信度检测目标 边界. Zigzag [38] 和Zhang等人 [24] 衡量图像中目标定位困难程度, 在训练过程中从易到难训练样本, 获得 更好的检测效果.虽然当下已提出许多基于多实例学习的弱监督目标检测算法, 验证了多实例检测网络与在线精细 化实例分类网络结合后的有效性, 但是在空间、类别和实例间丰富依赖关系上仍处于探索阶段.…”