Accepted by ICCV2023, Revisiting Foreground and Background Separation in Weakly-supervised Temporal Action Localization: A Clustering-based Approach
104
stars
3
forks
source link
关于ActivityNet v1.2和ActivityNet v1.3数据集训练的细节问题 #5
Open
mymuli opened 6 months ago
您好！ 我注意到你们组先后发表了一下几篇论文： AFPS：《Weakly supervised temporal action localization with actionness-guided false positive suppression》 CASE: 《Revisiting Foreground and Background Separation in Weakly-supervised Temporal Action Localization: A Clustering-based Approach》 AICL:《Actionness Inconsistency-Guided Contrastive Learning for Weakly-Supervised Temporal Action Localization》 其中，CASE模型和AICL分别在THUMOS14数据集上的AVG mAP(0.1:0.7)指标上的性能为46.2和46.4，但在ActivityNet v1.2数据集上可以分别达到27.9和29.9，远高于其他方法。
请问一下是否可以提供ActivityNet v1.2数据集的特征，训练参数和对应的细节呢？谢谢