Scalsol / mega.pytorch

Memory Enhanced Global-Local Aggregation for Video Object Detection, CVPR2020
Other
568 stars 115 forks source link

The results of RDN #55

Open xc-csc101 opened 4 years ago

xc-csc101 commented 4 years ago

Thanks for your impressive codebase. I try to reproduce the results of RDN, but the best result is 80.9 which is lower than 81.7 obviously. Shoud I modify any parameters or add any training tricks?

ZhijunHou commented 4 years ago

Hi, I just got 78.%. Have you modified any parameters setting?

xc-csc101 commented 4 years ago

Hi, I just got 78.%. Have you modified any parameters setting?

No, I just use default settings. According to my experimental results, the map is different every time because of random sampling, the avearge map I got is about 80.6

ZhijunHou commented 4 years ago

Thanks!

zhanghaoo commented 3 years ago

为什么你们都可以跑通啊!啊啊啊啊啊啊啊啊!

我的为什么出来start training就不动了呢?

b762927 commented 1 year ago

Hello, I want to ask the RDN how to train in two-stage, I find the network first-stage training code is the same the second。