mega.pytorch icon indicating copy to clipboard operation
mega.pytorch copied to clipboard

Memory Enhanced Global-Local Aggregation for Video Object Detection, CVPR2020

Results 65 mega.pytorch issues
Sort by recently updated
recently updated
newest added

If cloning the latest apex repo directly, line 19 of `apex/apex/transformer/amp/grad_scaler.py` used `torch.cuda.amp.GradScaler`, which requires torch version to be greater than 1.6 in order for `torch.cuda` to have `amp` module....

I followed install.md when running command python setup.py build develop my nvcc --version is cuda 10.1 I tried conda install pytorch==1.4.0 torchvision==0.5.0 cudatoolkit=10.1 -c pytorch conda install pytorch==1.5.0 torchvision==0.6.0 cudatoolkit=10.1...

![image](https://user-images.githubusercontent.com/74225912/115327726-02584a00-a1c2-11eb-91ef-e3fa94d850da.png)

I just don't know why we must use the DET datasets? The DET datasets just for riching the datasets or have the special meaning ?

在单GPU下进行消融实验,实验结果发现在没有memory的情况下mAP为83.18%,比完整的MEGA效果(83.18%mAP)还要好,这问题如何解决?? @Scalsol

看了您code中FGFA中的部分,发现了一个问题,在测试的时候,您的代码没有导入flownet的预训练模型,经过我试验得到的结果就是,对于一个相同的图像序列,我用您的demo中的FGFA方法进行检测,什么都不改变,检测两遍会得到不同的结果,结果差别比较大,但是在demo中加入checkpointer.load_flownet(),进行多遍检测,结果是一样的,希望您和使用者周知。