Attention-Based Dropout Layer for Weakly Supervised Object Localization

By Junsuk Choe, Hyunjung Shim. 2019

412 citations recorded by the source corpus.

Weakly Supervised Object Localization (WSOL) techniques learn the object location only using image-level labels, without location annotations. A common limitation for these techniques is that they cover only the most discriminative part of the object, not the entire object. To address this problem, we propose an Attention-based Dropout Layer (ADL), which utilizes the self-attention mechanism to process the feature maps of the model. The proposed method is composed of two key components: 1) hiding the most discriminative part from the model for capturing the integral extent of object, and 2) highlighting the informative region for improving the recognition power of the model. Based on extensive experiments, we demonstrate that the proposed method is effective to improve the accuracy of WSOL, achieving a new state-of-the-art localization accuracy in CUB-200-2011 dataset. We also show that the proposed method is much more efficient in terms of both parameter and computation overheads than existing techniques.

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

1. Lijuan Wang
2. Zhe Gan
3. Zicheng Liu
4. Jianfeng Wang
5. Zhengyuan Yang
6. Xiaowei Hu
7. Jianfeng Gao
8. Qibin Hou

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Ming-Hsuan Yang
3. Dacheng Tao
4. Trevor Darrell
5. Bernt Schiele
6. Xiaogang Wang
7. Andrew Zisserman
8. Jiashi Feng

Citations in the corpus

• CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features (2019, methodology): “CutMix also achieves comparable localization accuracies on CUB200-2011 and ImageNet, even when compared against the dedicated state-of-the-art WSOL methods [51, 32, 48, 49, 2] that focus on learning spatially dispersed representations.”
• CABNet: Category Attention Block for Imbalanced Diabetic Retinopathy Grading (2020, background): “Attention mechanisms are helpful to capture the fine-grained features in most computer vision tasks and have been widely used for image classification [21], [22], object localization [23], and semantic segmentation [24], [25].”
• LayerCAM: Exploring Hierarchical Class Activation Maps for Localization (2021, methodology): “The effectiveness of the attention methods [1], [2] or attention modules [23], [24], [25] benefit a lot of vision tasks [26], [27], [28], [29], [30], [31], [32].”
• Weakly Supervised Object Localization and Detection: A Survey (2021, positive): “4 Although there are other methods that use ResNet as the backbone [2], [20], [188], 3.”
• UniTAB: Unifying Text and Box Outputs for Grounded Vision-Language Modeling (2021, background): “The “GT-known” [57,75,76,15] setting aims to localize a given ground-truth class.”
• Aggregating Nested Transformers (2021, background): “NesT Hierarchy (Froward direction is 5 to 1) 192 , 384 , 768] and h = [3 , 6 , 12] for model T, S, and B , and k = [8 , 20 , 20] for model T, S, and B Architecture configuration.”
• Visual Attention Methods in Deep Learning: An In-Depth Survey (2022, methodology): “[129] proposed using the dropout layer to improve the localization accuracy through two steps: making the whole object location even by hiding the most discriminative part and attending over the whole area to improve the recognition performance.”
• Multi-branch and Multi-scale Attention Learning for Fine-Grained Visual Categorization (2020, background): “perimental results prove the effectiveness of these methods to improve object location accuracy. Our weakly supervised object location method can achieve higher location accuracy than ACOL[20] and ADL[21], etc. without adding trainable parameters, we will demonstrate it’s performance in the next sec”
• Sketch-Guided Object Localization in Natural Images (2020, background): “In [3,31], the authors proposed attention networks to generate attention scores on each of the object proposals, and an attention-based dropout layer for weakly supervised object localization.”
• Crossing the Format Boundary of Text and Boxes: Towards Unified Vision-Language Modeling (2021, background): “The first is the “GT-known” [12,50,67,68], where models know the ground-truth class to localize.”
• Comprehensive Attention Self-Distillation for Weakly-Supervised Object Detection (2020, positive): “[20, 21, 22, 23, 24, 25] re-weight features with spatial-wise attention maps.”
• ResizeMix: Mixing Data with Preserved Object Information and True Labels (2020, background): “Cutting- and Mixing- based Data Augmentation The goal of cutting augmentations is to make a network pay attention to the entire data like the dropout regularization [7, 18, 20, 46, 47].”