W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection

By Yongqiang Zhang, Yancheng Bai, Ming Ding, Yongqiang Li, Bernard Ghanem. 2018

130 citations recorded by the source corpus.

Weakly-supervised object detection has attracted much attention lately, since it does not require bounding box annotations for training. Although significant progress has also been made, there is still a large gap in performance between weakly-supervised and fully-supervised object detection. Recently, some works use pseudo ground-truths which are generated by a weakly-supervised detector to train a supervised detector. Such approaches incline to find the most representative parts of objects, and only seek one ground-truth box per class even though many same-class instances exist. To overcome these issues, we propose a weakly-supervised to fully-supervised framework, where a weakly-supervised detector is implemented using multiple instance learning. Then, we propose a pseudo ground-truth excavation (PGE) algorithm to find the pseudo ground-truth of each instance in the image. Moreover, the pseudo ground-truth adaptation (PGA) algorithm is designed to further refine the pseudo ground-truths from PGE. Finally, we use these pseudo ground-truths to train a fully-supervised detector. Extensive experiments on the challenging PASCAL VOC 2007 and 2012 benchmarks strongly demonstrate the effectiveness of our framework. We obtain 52.4% and 47.8% mAP on VOC2007 and VOC2012 respectively, a significant improvement over previous state-of-the-art methods.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Bridging Non Co-occurrence with Unlabeled In-the-wild Data for Incremental Object Detection (2021, neutral): “In the task of image-based object detection, object detectors are trained on large-scale datasets and then deployed on real-world applications [9, 11, 8, 24, 5, 15, 23, 17, 16, 1, 2, 39, 33, 34, 38, 36, 37, 35, 32].”
Weakly Supervised Object Localization and Detection: A Survey (2021, unknown): “Shen-CVPR-2018 [125] SSD CNN None ImageNet(tag label) SSD+Fast RCNN* MIL+GAN General objects Zhang-CVPR-2018 [215] Faster RCNN (VGG16) CNN None ImageNet(tag label) MIDN+Faster RCNN WSOD+Pseudo Ground-truth Mining+FOD General objects”
Dissimilarity Coefficient Based Weakly Supervised Object Detection (2018, methodology): “The weakly supervised detector employed in W2F models the annotation constraint using a fully factorized distribution.”
C-MIDN: Coupled Multiple Instance Detection Network With Segmentation Guidance for Weakly Supervised Object Detection (2019, methodology): “[43] propose a Weakly-Supervised to Fully-Supervised Framework(W2F) which use PGA and PGE to mine better pseudo ground truth from MIDN to train a fully-supervised detector.”
From Discriminant to Complete: Reinforcement Searching-Agent Learning for Weakly Supervised Object Detection (2020, background): “into the existing WSOD networks, such as [9] and [41]–[44], to discover more reliable target regions among the WSL process.”
Multi-task Generative Adversarial Network for Detecting Small Objects in the Wild (2020, methodology): “Similarly, the performance of object detection (Cai and Vasconcelos 2018; Bai et al. 2018b; Zhang et al. 2018b,c, 2019d) has been signiﬁcantly boosted due to richer appearance and spatial representations, which are learned by CNNs (Girshick et al. 2014) from large scale image datasets.”
Detecting 11K Classes: Large Scale Object Detection Without Fine-Grained Bounding Boxes (2019, background): “The most likely reason is that modern CNNs tend to focus on the most discriminative part of objects rather than the whole object [39].”