LocNet: Improving Localization Accuracy for Object Detection

By Spyros Gidaris, Nikos Komodakis. 2015

148 citations recorded by the source corpus.

We propose a novel object localization methodology with the purpose of boosting the localization accuracy of stateof-the-art object detection systems. Our model, given a search region, aims at returning the bounding box of an object of interest inside this region. To accomplish its goal, it relies on assigning conditional probabilities to each row and column of this region, where these probabilities provide useful information regarding the location of the boundaries of the object inside the search region and allow the accurate inference of the object bounding box under a simple probabilistic framework. For implementing our localization model, we make use of a convolutional neural network architecture that is properly adapted for this task, called LocNet. We show experimentally that LocNet achieves a very significant improvement on the mAP for high IoU thresholds on PASCAL VOC2007 test set and that it can be very easily coupled with recent stateof-the-art object detection systems, helping them to boost their performance. Finally, we demonstrate that our detection approach can achieve high detection accuracy even when it is given as input a set of sliding windows, thus proving that it is independent of box proposal 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

Wide Residual Networks (2016, methodology): “We also used WRN-34-2 to participate in COCO 2016 object detection challenge, using a combination of MultiPathNet [32] and LocNet [7].”
Cascade R-CNN: High Quality Object Detection and Instance Segmentation (2019, methodology): “[17], [18], [59] used a multi-stage procedure to generate accurate proposals, which are forwarded”
Cascade R-CNN: Delving Into High Quality Object Detection (2017, methodology): “[36, 12, 11] used a multi-stage procedure to generate accurate proposals, and forwarded them to an accurate model (e.”
Attend Refine Repeat: Active Box Proposal Generation via In-Out Localization (2016, methodology): “ihood of the in-out elements of B˜. A visual illustration of the above process through a few examples is provided in Fig.3(for further details about the LocNet model we refer the interested reader to [14]). We note that in contrast to the original LocNet model that is optimized to yield a different ”
Object Detection in 20 Years: A Survey (2019, unknown): “Besides, some researchers also tried to improve localization under a probabilistic inference framework [191].”
Recent Advances in Deep Learning for Object Detection (2019, background): “3 b LocNet [149] VGG-16 2016 ∼ 600 × 1000 78.”
Deep Regionlets for Object Detection (2017, unknown): “ike ResNet-101 [18]. Second, comparing with DP-FCN [32] and Deformable ROI Methods mAP@0.5 / @0.7(%) Faster R-CNN [36] 78.1 / 62.1 SSD [28] 76.8 / N/A DP-FCN [32] 78.1 / N/A ION [2] 79.4 / N/A LocNet [13] 78.4 / N/A Deformable ConvNet [6] 78.6 / 63.3 Deformale ROI Pooling [6] 78.3 / 66.6 Deformable ”
Localization Distillation for Object Detection (2022, background): “Object localization is a fundamental issue in object detection [18], [33], [44], [64], [65], [73], [74], [75], [80], [93].”
HyperFace: A Deep Multi-Task Learning Framework for Face Detection, Landmark Localization, Pose Estimation, and Gender Recognition (2019, methodology): “Instead of using the probabilistic framework [14], we solve the above mentioned issues in an iterative way using the predicted landmarks.”
Hierarchical and Robust Convolutional Neural Network for Very High-Resolution Remote Sensing Object Detection (2019, background): “However, it has been proven that the localization needs more delicate inputs and local features are more suitable than global features [20].”
Standard detectors aren't (currently) fooled by physical adversarial stop signs (2017, methodology): “While there are post-processing methods to improve boxes [6], this tension is fundamental (for example, see figure 2 and 3).”
ContextLocNet: Context-Aware Deep Network Models for Weakly Supervised Localization (2016, background): “com/ShaoqingRen/SPP_net 3 This choice for the frame parameters follows [39, 40], and the ratio is kept same for both context and frame pooling types.”
Accurate Single Stage Detector Using Recurrent Rolling Convolution (2017, background): “The motivation of [7] is similar to ours, but it does not consider contextual information by using recurrent architecture.”
SNIPER: Efficient Multi-Scale Training (2018, methodology): “We do not perform flipping [36], iterative bounding box regression [7] or mask tightening [19].”
Focal and Efficient IOU Loss for Accurate Bounding Box Regression (2021, methodology): “Till now, BBR has been used on most of the recent detection methods [5, 8, 21, 24, 1, 16, 6].”
BSN: Boundary Sensitive Network for Temporal Action Proposal Generation (2018, methodology): “Boundary probabilities are also adopted in LocNet [22] for revising the horizontal and vertical boundaries of existing proposals.”
Seesaw Loss for Long-Tailed Instance Segmentation (2020, background): “There are also attempts to apply cascade architecture [1, 10, 22, 35, 49] to reﬁne the bounding boxes’ predictions progressively.”