Mutual Enhancement for Detection of Multiple Logos in Sports Videos

Liao, Yuan; Lu, Xiaofeng; Zhang, Chengcui; Wang, Yongtao; Tang, Zhi

doi:10.1109/iccv.2017.519

Cited by 27 publications

(13 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We compare the performance of DETR-FP against a strong Faster-RCNN baseline in the task of Logo detection concretely in the QMUL-Openlogo benchmark [12]. The dataset is comprised of 27,083 images from 352 logo classes, built by aggregating and refining seven other logo datasets [58], [59], [67]- [71]. We use weights from ImageNet and MS-COCO [11] in different experiments for both architectures.…”

Section: Methodsmentioning

confidence: 99%

Logo Detection With No Priors

Velazquez

Gonfaus²,

Rodríguez³

et al. 2021

IEEE Access

View full text Add to dashboard Cite

In recent years, top referred methods on object detection like R-CNN have implemented this task as a combination of proposal region generation and supervised classification on the proposed bounding boxes. Although this pipeline has allowed to achieve state-of-the-art results in multiple datasets, it has inherent limitations that has converted object detection into a very complex and inefficient task in computational terms. In contrast to considering this standard strategy, in this paper we enhance Detection Transformers (DETR) which originally implements object detection as a set-prediction problem directly in an end-to-end fully differentiable pipeline without requiring priors. In particular, we incorporate Feature Pyramids (FP) to the DETR architecture and demonstrate the effectiveness of the resulting DETR-FP approach on improving logo detection results, in essence thanks to the correct detection of small logos. So, without requiring any domain specific prior to be fed to the model, DETR-FP obtains competitive results on the OpenLogo and MS-COCO datasets, when compared to a Faster R-CNN baseline which strongly depends on hand-designed priors.

show abstract

Section: Methodsmentioning

confidence: 99%

Logo Detection With No Priors

Velazquez

Gonfaus²,

Rodríguez³

et al. 2021

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Classes Images Availability BelgaLogos (Joly and Buisson, 2009) 37 1,321 FlickrLogos-27 (Kalantidis et al, 2011) 27 810 FlickrLogos-32 (Romberg et al, 2011) 32 2,240 Logo32plus (Bianco et al, 2017) 32 7,830 Logo-In-The-Wild (Tzk. et al, 2018) 1196 9,393 SportsLogo (Liao et al, 2017) 20 1,978 MICC-Logos (Sahbi et al, 2012) 13 720 LOGO-NET (Hoi et al, 2015) 160 73,414 OpenLogo (Su et al, 2018) 352 27,083 been a number of logo detection datasets developed in the literature (Joly and Buisson, 2009;Kalantidis et al, 2011;Romberg et al, 2011;Bianco et al, 2017;Tzk. et al, 2018;Liao et al, 2017;Sahbi et al, 2012;Hoi et al, 2015;Su et al, 2018) (Table 1).…”

Section: Datasetmentioning

confidence: 99%

“…et al, 2018) 1196 9,393 SportsLogo (Liao et al, 2017) 20 1,978 MICC-Logos (Sahbi et al, 2012) 13 720 LOGO-NET (Hoi et al, 2015) 160 73,414 OpenLogo (Su et al, 2018) 352 27,083 been a number of logo detection datasets developed in the literature (Joly and Buisson, 2009;Kalantidis et al, 2011;Romberg et al, 2011;Bianco et al, 2017;Tzk. et al, 2018;Liao et al, 2017;Sahbi et al, 2012;Hoi et al, 2015;Su et al, 2018) (Table 1). However, their scaling ability is limited in terms of both class and image, due to the high cost of collecting and labelling in-the-wild logo images.…”

Section: Datasetmentioning

confidence: 99%

“…Due to remarkable success of deep learning (Nanni et al, 2017), recent state-of-the-art logo detection methods leverage generic object detection networks (Girshick, 2015;Ren et al, 2015;Redmon and Farhadi, 2017). For instance, Iandola et al (2015) and Liao et al (2017) exploit Fast R-CNN. Later on, Faster R-CNN are often selected (Hoi et al, 2015;Su et al, 2017b) due to the superior efficiency and performance.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Multi-perspective cross-class domain adaptation for open logo detection

Gong

Zhu³

2021

Computer Vision and Image Understanding

View full text Add to dashboard Cite

“…1 (c)). Existing logo detection methods typically consider a small number of logo classes with the need for large sized training data annotated with object bounding boxes [2,14,15,19,20,24,31,32,33]. Whilst this controlled setting allows for a straightforward adoption of the state-of-the-art general object detection models [6,28,30], it is unscalable to dynamic real-world logo detection applications where more new logo classes become of interest during model deployment, with the availability of only their clean design images (Fig.…”

Section: Introductionmentioning

confidence: 99%

Open Logo Detection Challenge

Su,

Zhu,

Gong

2018

Preprint

View full text Add to dashboard Cite

Existing logo detection benchmarks consider artificial deployment scenarios by assuming that large training data with fine-grained bounding box annotations for each class are available for model training. Such assumptions are often invalid in realistic logo detection scenarios where new logo classes come progressively and require to be detected with little or none budget for exhaustively labelling fine-grained training data for every new class. Existing benchmarks are thus unable to evaluate the true performance of a logo detection method in realistic and open deployments. In this work, we introduce a more realistic and challenging logo detection setting, called Open Logo Detection. Specifically, this new setting assumes fine-grained labelling only on a small proportion of logo classes whilst the remaining classes have no labelled training data to simulate the open deployment. We further create an open logo detection benchmark, called QMUL-OpenLogo, to promote the investigation of this new challenge. QMUL-OpenLogo contains 27,083 images from 352 logo classes, built by aggregating/refining 7 existing datasets and establishing an open logo detection evaluation protocol. To address this challenge, we propose a Context Adversarial Learning (CAL) approach to synthesising training data with coherent logo instance appearance against diverse background context for enabling more effective optimisation of contemporary deep learning detection models. Experiments show the performance advantage of CAL over existing state-of-the-art alternative methods on the more realistic and challenging open logo benchmark. The QMUL-OpenLogo benchmark is publicly available at https://qmul-openlogo.github.io/.

show abstract

Mutual Enhancement for Detection of Multiple Logos in Sports Videos

Cited by 27 publications

References 26 publications

Logo Detection With No Priors

Logo Detection With No Priors

Multi-perspective cross-class domain adaptation for open logo detection

Open Logo Detection Challenge

Contact Info

Product

Resources

About