search

haifangong / CMSA-MTPT-4-MedicalVQA

[ICMR'21, Best Poster Paper Award] Medical Visual Question Answering with Multi-task Pre-training and Cross-modal Self-attention
MIT License
33 stars 2 forks source link

What is the extraction code of the extended data set? #4

Closed jiuge-sun closed 2 years ago

haifangong commented 2 years ago

We have released the link of the dataset on Baidu Drive and Google Drive in the newest version of the README file.

rhyhck commented 11 months ago

No description provided.

Hello, may I ask if the results you reproduced are the same as the results of paper? My results are as follows, which differs greatly from the paper. I have not made any modifications to the code. {"CLOSED": {"count": 185.0, "real": 180.0, "true": 135.0, "real_percent": 0.0, "score": 0.7297297297297297, "score_percent": 73.0}, "OPEN": {"count": 123.0, "real": 90.0, "true": 45.0, "real_percent": 0.0, "score": 0.365853658536, "score_percent": 36.6}, ALL ": {" count ": 308.0," real ": 270.0," true ": 180.0," real_percent ": 0.0," score ": 0.5844155844155844," score_percent ": 58.4}} I'm not sure if I missed the parameters that need to be loaded?