guojingjuan的博客

We use a two-stage Faster R-CNN style object detection framework [12] and leverage an SE-ResNeXt or SENet [4] model as the backbone feature extractor. To increase the global context information in t...

https://mp.weixin.qq.com/s/cP2kM553XyRsUZ5xJ8kAZA 以下为百度视觉团队技术方案解读：存在挑战与传统的检测数据集合相比，该赛事除了数据规模大、更真实之外，还存在一系列的挑战。具体来说，主要集中在以下三个方面： 数据分布不均衡：最少的类别框选只有 14 个，而最多的类别框选超过了 140w，数据分布严重不均衡。 类别框数量分...

Dual Refinement Network for Single-Shot Object DetectionPooling Pyramid Network for Object DetectionCornerNet: Detecting Objects as Paired Keypoints（eccv2018）Tiny-DSOD: Lightweight Object Detect...

  multibox_loss_param {    loc_loss_type: SMOOTH_L1    conf_loss_type: SOFTMAX    loc_weight: 1.0    num_classes: 21    share_location: true    match_type: PER_PREDICTION    overlap_threshold: ...

http://openaccess.thecvf.com/ECCV2018_search.py 1.License Plate Detection and Recognition in Unconstrained Scenarios [pdf] 2.Bidirectional Feature Pyramid Network with Recurrent Attention Residu...

https://www.jiqizhixin.com/articles/2018-08-03-19面向目标检测任务的深度特征金字塔再组合Deep Feature Pyramid Reconfiguration for Object Detection该论文由腾讯AI Lab与清华大学合作完成。目前最好的目标检测器大多通过特征金字塔来学习多尺度表示从而取得更高的检测精度。然而，当前特征金...

Code is available at https://github.com/ruinmessi/RFBNet.