Outputs of evaluation(LSE-C, LSE-D) consist of four values for each, not one for each?

[justin4ai]

Fellas, as you can see in the image below, it seems the evaluation code(sh calculate_scores_real_videos.sh PATH/TO/DATASET) outputs 8 values(4 x 2), which is different from what I expected - 2 values(one for LSE-C and another for LSE-D).

Can anyone let me know how I should interpret these numbers?

Thanks,
Junyeong Ahn

[Crestina2001]

Hi, I am also working on the project, but cannot find an evaluation code. Do you implement the calculate_scores_real_videos.sh by yourself?

[justin4ai]

Hi, I am also working on the project, but cannot find an evaluation code. Do you implement the calculate_scores_real_videos.sh by yourself?

No I didn't. Please just follow the instruction line by line, then the file will be there. It is from SyncNet official repository.

[Crestina2001]

Hi, I am also working on the project, but cannot find an evaluation code. Do you implement the calculate_scores_real_videos.sh by yourself?

No I didn't. Please just follow the instruction line by line, then the file will be there. It is from SyncNet official repository.

I have found the evaluation script specified in the evaluation folder. Now looking into how to use it.

[justin4ai]

Hi, I am also working on the project, but cannot find an evaluation code. Do you implement the calculate_scores_real_videos.sh by yourself?

No I didn't. Please just follow the instruction line by line, then the file will be there. It is from SyncNet official repository.

I have found the evaluation script specified in the evaluation folder. Now looking into how to use it.

All good! And please also let me know if you get to know what the resulting 8 numbers mean now and then!

[Crestina2001]

Hi, I am also working on the project, but cannot find an evaluation code. Do you implement the calculate_scores_real_videos.sh by yourself?

No I didn't. Please just follow the instruction line by line, then the file will be there. It is from SyncNet official repository.

I have found the evaluation script specified in the evaluation folder. Now looking into how to use it.

All good! And please also let me know if you get to know what the resulting 8 numbers mean now and then!

Hi. You may change the 38th line in the calculate_scores_real_videos.py into:

for idx, fname in enumerate(flist):
offset, conf, dist = s.evaluate(opt,videofile=fname)
print(f"File: {fname}, Sync Error: {dist}, Confidence: {conf}, Offset: {offset}")

In this way, the result will be more understandable.

[justin4ai]

Hi, I am also working on the project, but cannot find an evaluation code. Do you implement the calculate_scores_real_videos.sh by yourself?

No I didn't. Please just follow the instruction line by line, then the file will be there. It is from SyncNet official repository.

I have found the evaluation script specified in the evaluation folder. Now looking into how to use it.

All good! And please also let me know if you get to know what the resulting 8 numbers mean now and then!

Hi. You may change the 38th line in the calculate_scores_real_videos.py into:

for idx, fname in enumerate(flist): offset, conf, dist = s.evaluate(opt,videofile=fname) print(f"File: {fname}, Sync Error: {dist}, Confidence: {conf}, Offset: {offset}")

In this way, the result will be more understandable.

Thanks mate! Still I don't get why I have four rows tho.

[justin4ai]

@Crestina2001 Mate, do you have any idea about why there are four numbers for each score? I still don't get why it is that.