Jian
Jian
your result is not good Area Under the AR vs AN curve should be about 68%.
Because of randomness, you can ignore the difference after the decimal point in the result.
Hi, semchan we use some tricks like the multi-scale ensemble.
Fast
@ideaRunner After you use TSN to extract video features, the features length is different. You must reshape these features by fitting and interpolation functions. 
Hi @pengxiaoxiao With these inference codes，you can repeat the results in our paper.
@pengxiaoxiao check your tensorflow version
Hi @gtgtgt1117 对于100长特征而言，proposals的数量是固定的，有100*100/2，PFG layer其实是生成这些个proposals的特征，这些个proposals肯定是有重叠的，靠后面的两个分支进行打分和后处理去重。
@gtgtgt1117 是的，一般还需要卡一个分数阈值来保留。