Improving Object Localization with Fitness NMS and Bounded IoU Loss

By Lachlan Tychsen-Smith, Lars Petersson. 2017

194 citations recorded by the source corpus.

We demonstrate that many detection methods are designed to identify only a sufficently accurate bounding box, rather than the best available one. To address this issue we propose a simple and fast modification to the existing methods called Fitness NMS. This method is tested with the DeNet model and obtains a significantly improved MAP at greater localization accuracies without a loss in evaluation rate, and can be used in conjunction with Soft NMS for additional improvements. Next we derive a novel bounding box regression loss based on a set of IoU upper bounds that better matches the goal of IoU maximization while still providing good convergence properties. Following these novelties we investigate RoI clustering schemes for improving evaluation rates for the DeNet wide model variants and provide an analysis of localization performance at various input image dimensions. We obtain a MAP of 33.6%@79Hz and 41.8%@5Hz for MSCOCO and a Titan X (Maxwell). Source code available from: https://github.com/lachlants/denet

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 the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (2019, neutral): “9 FitnessNMS [57] trainval DeNet-101 39.”
Learning Open-World Object Proposals without Learning to Classify (2021, background): “Several works [15], [16], [14], [36] recalibrate the final detection confidence by using both localization and classification subnets.”
Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection (2020, methodology): “These two factors result in a gap between training and test, and would potentially degrade the detection performance, e.g., negative instances with randomly high-quality scores could rank in front of positive examples with lower quality prediction during NMS.”
CornerNet: Detecting Objects as Paired Keypoints (2018, neutral): “1 - - - - - Fitness-NMS (Tychsen-Smith and Petersson, 2017b) ResNet-101 41.”
MMDetection: Open MMLab Detection Toolbox and Benchmark (2019, background): “Recently, there are more regression losses proposed, e.g ., Bounded IoU Loss [33], IoU Loss [32], GIoU Loss [28], Balanced L1 Loss [23].”
Mask Scoring R-CNN (2019, methodology): “[36] proposed Fitness NMS that corrected the detection score using the IoU between the detected bounding boxes and their ground truth.”
Region Proposal by Guided Anchoring (2019, methodology): ““RPN+Focal Loss” and “RPN+Bounded IoU Loss” denotes adopting focal loss [19] and bounded IoU Loss [29] to RPN by substituting binary cross-entropy loss and smooth l1 loss, respectively.”
PIoU Loss: Towards Accurate Oriented Object Detection in Complex Environments (2020, background): “[37] introduce a novel bounding box regression loss based on a set of IoU upper bounds.”
High-Resolution Representations for Labeling Pixels and Regions (2019, unknown): “FitnessNMS [96] ResNet-101 768 - 39.”
Learning to Rank Proposals for Object Detection (2019, background): “Fitness NMS [34] introduces the localization information of bounding box into ranking conﬁdence.”
Enhancing Geometric Factors in Model Learning and Inference for Object Detection and Instance Segmentation (2020, background): “To ameliorate the training stability, Bounded-IoU loss [35] introduces the upper bound of IoU.”
Object Detection in 20 Years: A Survey (2019, unknown): “LearnNMS [110], RelationNet [93], Learn2Rank [111], SoftNMS [112], FitnessNMS [113], SofterNMS [114], AdaptiveNMS [115], DIoUNMS [107],”