PDVC icon indicating copy to clipboard operation
PDVC copied to clipboard

Running PDVC on Your Own Videos predict json file

Open Baiiiiiiiiii opened this issue 1 year ago • 2 comments

Hi, after I run the command

video_folder=visualization/videos output_folder=visualization/output pdvc_model_path=save/anet_tsp_pdvc/model-best.pth output_language=en bash test_and_visualize.sh $video_folder $output_folder $pdvc_model_path $output_language

I got a JSON file with the format as below, image

I am curious about the meaning of proposal_score and sentence_score. Could you give me some explanation? I want to use these two metrics to select better captioning sentences .

thank you!

Baiiiiiiiiii avatar Dec 26 '23 17:12 Baiiiiiiiiii

The proposal score is the logit score of foreground, which is predict by a linear layer (before signmoid activation) with two nodes representing background and foreground. The sentence score is the log of perplexity of the generated sentence (multiplication of per-word confidence)

ttengwang avatar Jan 11 '24 17:01 ttengwang

嗨,在我运行命令之后> 嗨,在我运行命令之后>> video_folder=visualization/videos output_folder=visualization/output pdvc_model_path=save/anetsp_p_pdvc/model-best.pth output_language=en bash test_and_visualize.sh $video_folder $output_folder $pdvc_model_path $output_language>> 我得到了一个格式如下的JSON文件, ![图像](https://private-user-images.githubusercontent.com/94792611/292899440-fa8ebf3e-8e6d-4217-9fa0-6c1b37dd115c.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MTY1NDIyMzIsIm5iI6MTcxNjU0MTkzMiwicGF0aCI6Ii85NDc5MjYxMS8yOTI4OTk0NDAtZm4Zm2UtOGU2ZC00MjE3LTlmYTAtNmMmMmMxYjM3ZGQxM3ZJFzI6w4bSWghs_UJyoNsh-Y3j2uON1Uf37of61KgDUo)>> 我对possion_score和sentente_score的含义感到好奇。你能给我一些解释吗?我想用这两个指标来选择更好的字幕句子。>> 谢谢你!>> video_folder=visualization/videos output_folder=visualization/output pdvc_model_path=save/anetsp_p_pdvc/model-best.pth output_language=en bash test_and_visualize.sh $video_folder $output_folder $pdvc_model_path $output_language>> 我得到了一个格式如下的JSON文件, ![图像](https://private-user-images.githubusercontent.com/94792611/292899440-fa8ebf3e-8e6d-4217-9fa0-6c1b37dd115c.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MTY1NDIyMzIsIm5iI6MTcxNjU0MTkzMiwicGF0aCI6Ii85NDc5MjYxMS8yOTI4OTk0NDAtZm4Zm2UtOGU2ZC00MjE3LTlmYTAtNmMmMmMxYjM3ZGQxM3ZJFzI6w4bSWghs_UJyoNsh-Y3j2uON1Uf37of61KgDUo)>> 我对possion_score和sentente_score的含义感到好奇。你能给我一些解释吗?我想用这两个指标来选择更好的字幕句子。>> 谢谢你! 你好,我在按照redme.rd配置环境的时候,出现一个问题,希望您能帮我看一下:我在执行 sh make.sh 命令时,出现以下报错:'nvalid command name 'install 我找了很多博客,都没能解决这个问题,您能否提供一些帮助,谢谢您!

Kebuze avatar May 24 '24 09:05 Kebuze