fromlimbo

Results 10 comments of fromlimbo

> @yjxiong about hmdb51 dataset, train:val=4:1, and when i test the val, Top1=60% hi,I was also testing my own dataset, so HOW did you transform you data meet the input...

> > > @yjxiong about hmdb51 dataset, train:val=4:1, and when i test the val, Top1=60% > > > > > > hi,I was also testing my own dataset, so HOW...

> @fromlimbo 你的数据集中视频时常是多少,作者的是10秒左右,大概在这个范围的分类结果会好,再调学习率什么的吧,调参玄学,感觉没什么规律~~~ 按照他的论文来的,本来是不到100帧,循环补全到300帧,10秒左右。然后我发现训练的时候weight bias是随机初始化的,没有用到pretrain的model。你这边训练用了pretrain的weight吗?

> @fromlimbo 你的数据集中视频时常是多少,作者的是10秒左右,大概在这个范围的分类结果会好,再调学习率什么的吧,调参玄学,感觉没什么规律~~~ 我这边从头训练的好了,top1准确率81.67%,top5 98%,应该还算可以了

> @yanqian123 @fromlimbo 我想请问通过openpose.py可以将多个帧的.json转换为video.json和我运行demo显示到/openpose_estimation/data/video.json的区别是什么? 没啥区别

> > > > > @yjxiong about hmdb51 dataset, train:val=4:1, and when i test the val, Top1=60% > > > > > > > > > > > > hi,I...

> > > > > @yjxiong about hmdb51 dataset, train:val=4:1, and when i test the val, Top1=60% > > > > > > > > > > > > hi,I...

@metropt is right. The FindCaffe.cmake file made some mistakes. Maybe our caffe is not the same version. The Caffe_DIR is ~/xx/caffe/distribute.(If your distribute file is empty,you shoud do "make distribute"...

实测下来流式首响基本都要800ms以上,不知道150ms怎么做的