SMVL active learning results.

Detection model based on FR-CNN [3] and Res101 [4] architecture. Comparison baselines includes DTAL, random selection, entropy ranking, DeepAL [1] and MOAID [2]. The active learning labelling batch size range from 20 to 100.

Fully supervised training results

DTAL Selection Metric Algorithm Codes Download

Matlab Codes

This is our original codes used in [5]. The codes are based on Matlab. Please note that the codes are in a coarse state, publicated as reference for reproducing the work in [5].

[1]. Hamed H Aghdam, Abel Gonzalez-Garcia, Joost van de Weijer and Antonio M Lopez. Active learning for deep detectin neural networks. ICCV 2019.

[2]. Tianning Yuan, Fang Wan, Mengying Fu, Jianzhuang Liu, Songcen Xu, Xiangyang Ji and Qixiang Ye. Multiple instance active learning for object detection. CVPR 2021.

[3]. Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real time object detection with region proposal networks. NIPS 2015.

[4]. Kaiming He, Xiangyu Zhang, Shaoqing Ren and Jian Sun. Deep residual learning for image recognition. CVPR 2016.

[5]. A Deep Transfer Active Learning Approach to Logo Detection in Sports Video. In Proc.