welix

@LuoweiZhou The difference of the scores are caused by the dataset split. In 2021 winter, I could not access the original test set annotations, so split the val set with...

I found how to calculate ngram and save a .p file here. https://github.com/ruotianluo/self-critical.pytorch/blob/master/scripts/prepro_ngrams.py The target dataset of this script is MSCOCO, but it may be helpful to you.

Thank you for merging the pull request. Due to the memory limit in my environment, I rewrite IM_BATCH_SIZE 60 to 24. Other experimental settings are the same to the original...

I have the same question, anyone solved?

@ZihaoZheng98 and @ermik Hi, thank you for your interest of our paper. I am a first author of the paper r-FG-BB dataset. All images and recipes in the r-FG-BB dataset...

The annotations of r-FG-BB dataset are grounding connection between recipe and image contents. Thus, metadata that I can publish is only "datasize", "created day", "format (json? pkl?)" ...etc, they are...

same issue. @jayleicn

@XiaohuJoshua @wjun0830 Hi, I also found that clip_len has a impact on the performance across multiple models: Moment DETR, QD-DETR, EaTR, UVCOM, and CGDETR. When conducting experiments on Charades-STA, please...

@Breakthrough Thank you for quick and detailed response! I understand how the values are determined. > The config file/command line parser handles this by storing the value as a string...

> Very cool - I've wanted to do [something similar for a while](https://github.com/Breakthrough/SceneStats), but haven't had time to pursue it. It looks like you have made a lot of progress,...