search

ControlNet / LAV-DF

[CVIU] Glitch in the Matrix: A Large Scale Benchmark for Content Driven Audio-Visual Forgery Detection and Localization
https://www.sciencedirect.com/science/article/pii/S1077314223001984
Other
73 stars 10 forks source link

visualize the result #15

Closed pradeepkc11 closed 10 months ago

pradeepkc11 commented 10 months ago

I am able to run the evaluate.py file with given checkpoint value.After this it created a output folder which contains the result folder with .csv files. can you please tell me how to use this files to get the output in Videos

ControlNet commented 10 months ago

Hi, the output folder contains the boundary map output for each video file. So each csv file is the inference results of corresponding video. The columns in the CSV is start, end and score. In evaluate.py, after they're generated, the soft non-maximumm suppression is applied to these CSV files and generate an aggregated JSON file located in outputs/results, which can be used for calculate the metrics.

pradeepkc11 commented 10 months ago

which script can be used to predict the fake or real for unseen video data, because their is no trained model is available to use it. I wanted to test the model on unseen data to know if it works for it and also which python script should i use to predict the unseen data

ControlNet commented 10 months ago

Currently, the unseen video visualizations are not implemented, but it's easy to achieve. I suggest loading the CSV file, apply soft nms using post_process.soft_nms and then visualized by yourself.

pradeepkc11 commented 10 months ago

ok thanks for the response and guidance