Learning Spatial Fusion for Single-Shot Object Detection

Liu, Songtao; Huang, Di; Wang, Yunlong

doi:10.48550/arxiv.1911.09516

Cited by 199 publications

(237 citation statements)

References 45 publications

Supporting

Mentioning

168

Contrasting

Unclassified

Order By: Relevance

“…When training the R-CNN module, the IoU threshold for pos/neg division is changed to 0.5. IoU based label assignment is proved effective and soon been adopted by many Faster R-CNN's variants like [2,12,20,42,26,49,37], as well as many one-stage detectors like [31,32,25,27,23,21].…”

Section: Fixed Label Assignmentmentioning

confidence: 99%

OTA: Optimal Transport Assignment for Object Detection

Ge¹,

Liu²,

Li³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Recent advances in label assignment in object detection mainly seek to independently define positive/negative training samples for each ground-truth (gt) object. In this paper, we innovatively revisit the label assignment from a global perspective and propose to formulate the assigning procedure as an Optimal Transport (OT) problem -a well-studied topic in Optimization Theory. Concretely, we define the unit transportation cost between each demander (anchor) and supplier (gt) pair as the weighted summation of their classification and regression losses. After formulation, finding the best assignment solution is converted to solve the optimal transport plan at minimal transportation costs, which can be solved via Sinkhorn-Knopp Iteration. On COCO, a single FCOS-ResNet-50 detector equipped with Optimal Transport Assignment (OTA) can reach 40.7% mAP under 1× scheduler, outperforming all other existing assigning methods. Extensive experiments conducted on COCO and CrowdHuman further validate the effectiveness of our proposed OTA, especially its superiority in crowd scenarios. The code is available at https: //github.com/Megvii-BaseDetection/OTA.

show abstract

Section: Fixed Label Assignmentmentioning

confidence: 99%

OTA: Optimal Transport Assignment for Object Detection

Ge¹,

Liu²,

Li³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…T is the number of previous features used for aggregation. Similar to [24], w is predicted by two convolution layers followed by softmax function. We find that in experiment the weighted summation is slightly better than average summation.…”

Section: Motion-guided Feature Warpermentioning

confidence: 99%

Track to Detect and Segment: An Online Multi-Object Tracker

Wu¹,

Cao²,

Song³

et al. 2021

Preprint

View full text Add to dashboard Cite

Most online multi-object trackers perform object detection stand-alone in a neural net without any input from tracking. In this paper, we present a new online joint detection and tracking model, TraDeS (TRAck to DEtect and Segment), exploiting tracking clues to assist detection end-to-end. TraDeS infers object tracking offset by a cost volume, which is used to propagate previous object features for improving current object detection and segmentation. Effectiveness and superiority of TraDeS are shown on 4 datasets, including MOT (2D tracking), nuScenes (3D tracking), MOTS and Youtube-VIS (instance segmentation tracking). Project page: https://jialianwu.com/ projects/TraDeS.html.

show abstract

“…Since SSD [6] and FPN [7] propose to detect objects with different sizes at different scale levels, there are numerous methods to extract better aligned features for different scale levels. PaNet [8], BiFpn [9] and ASFF [10] merge scale information in a deeper and more complex manner, the performance gap between two tasks is mitigated as each scale level includes more comprehensive information from other scale levels. Recursive-Fpn [11] uses ASPP [12] structure to merge information across scale levels, also mitigating the performance gap between two tasks.…”

Section: Scale Misalignmentmentioning

confidence: 99%

Toward Minimal Misalignment at Minimal Cost in One-Stage and Anchor-Free Object Detection

Hao¹,

Liu²,

Wang³

et al. 2021

Preprint

View full text Add to dashboard Cite

Common object detection models consist of classification and regression branches, due to different task drivers, these two branches have different sensibility to the features from the same scale levels and the same spatial locations. The pointbased prediction method, which is based on the assumption that the point with high classification confidence has the high regression quality, leads to the misalignment problem. According to our analyses, the problem is further composed of scale misalignment and spatial misalignment specifically. We aim to resolve the phenomenon at minimal cost-a minor adjustment of the head network and a new label assignment method replacing the rigid one. Our experiments show that, compared to the baseline FCOS, a one-stage and anchor-free object detection model, our model consistently get around 3 AP improvement with different backbones, demonstrating both simplicity and efficiency of our method. Our code will be available at https://github.com/HaoGood/MMMC.

show abstract

Learning Spatial Fusion for Single-Shot Object Detection

Cited by 199 publications

References 45 publications

OTA: Optimal Transport Assignment for Object Detection

OTA: Optimal Transport Assignment for Object Detection

Track to Detect and Segment: An Online Multi-Object Tracker

Toward Minimal Misalignment at Minimal Cost in One-Stage and Anchor-Free Object Detection

Contact Info

Product

Resources

About