SCAN
SCAN copied to clipboard
PyTorch source code for "Stacked Cross Attention for Image-Text Matching" (ECCV 2018)
Hi, I am interested in understanding your code. Can you provide a pre-trained model for microsoft coco ? Thank you for your help and for your contribution. Kent
Hello @kuanghuei. I read your paper and the "Data pre-processing" section carefully, but I could not find how to produce my vocab.json file for my custom dataset. Could it be...
FileNotFoundError: [Errno 2] No such file or directory: 'G://github下载/GSMN-master/data/coco_precomp/train_precaps_stan.txt' run file lack train_precaps_stan.txt'
I want to ask a question. For coco I use precomp dataset to classfication,but the performence is not good,I use full connection to make the image dimension to 80,and use...
Hello @kuanghuei , First congrats for your paper and repo. Since your repo uses the precomputed visual features from the Bottom-up model, but sadly if somebody wants to use your...
How the attention visualization in the paper is realized?
Hi I want to run evaluation with following command evaluation.evalrank("$RUN_PATH/coco_scan/model_best.pth.tar", data_path="$DATA_PATH", split="test") where can i download this pretrained model "model_best.pth.tar"? Thanks a lot!
hello, kuanghui I have read you paper and code carefully and I am interested in your approach, I found that there are the results of (i2t Avg + t2i LSE)method...
https://github.com/kuanghuei/SCAN/blob/01812aa98e2ebe39695c8906589b6fe66b2a0d6e/util/convert_data.py#L58
The f30k dataset renumbered the pictures. Could you share the image number corresponding to the f30k dataset. I can't find the picture corresponding to the text in f30k. Thank you...