ruiwang2021 / mvd Goto Github PK

Hi there,

I was just curious if you had any scripts that would allow us to run inference on our own videos, if so, could you please provide it?
If not, can you advice us on how to go about it?

Thanks!

want to have a try! other we will use videoMAE instead if it still need a time.

Thanks

Just wanted to say that I really enjoyed your work and was curious about your timetable to release the code?

作者你好，
我这里想使用一下你在SSV2数据集上的预训练模型，请问是否方便提供一下？

Hello,

Would it be possible to release the checkpoints for Something Something V2 (at least for Vit-Large)? Thank you!

Could you provide an example of how to set distribute parameters to run on 1 node with 2 GPUs:
OMP_NUM_THREADS=1 python -m torch.distributed.launch --nproc_per_node=${GPUS}
--master_port ${MASTER_PORT} --nnodes=${NODE_COUNT}
--node_rank=${RANK} --master_addr=${MASTER_ADDR} \

作者你好，我注意到你在DATASET.md中提到SSv2使用的测试和验证集是相同的，请问下你论文中的精度的测试方式是和VideoMAE中保持一致的吗？但是VideoMAE作者在其repo中提供的test和val split实际上都是SSv2官方提供的验证集（一共24777个样本），官方测试集则是另外一个有着27157个样本的版本。所以我现在有些困惑，实际论文中MVD，VideoMAE以及之前的方法使用的测试集是一致的吗？是都采用的SSv2官方验证集还是测试集呢？期待你的回复，谢谢！

Thank you for making the code publicly available. When will pre-trained models be released?

Here should we multiply num_sample variable also in calculation of total batch size ?