AnasEmad11 / C2FPL

A Coarse-to-Fine Pseudo-Labeling (C2FPL) Framework for Unsupervised Video Anomaly Detection
11 stars 3 forks source link

How do I get the file "nalist.npy"? #3

Open snowy199 opened 5 months ago

snowy199 commented 5 months ago

nalist = np.load("nalist.npy") #UCF

nalist = np.load("nalist_XD.npy") #XD

These two files are not in the project, how do I get these two files?

Daan-UT commented 4 months ago

Did you ever figure this out?

AnasEmad11 commented 4 months ago

Hello,

Sorry for the delay, Please find attached the nalist.npy for UCF-Crime.


From: Daan-UT @.> Sent: Thursday, May 16, 2024 5:22 PM To: AnasEmad11/C2FPL @.> Cc: Subscribed @.***> Subject: Re: [AnasEmad11/C2FPL] How do I get the file "nalist.npy"? (Issue #3)

Did you ever figure this out?

— Reply to this email directly, view it on GitHubhttps://github.com/AnasEmad11/C2FPL/issues/3#issuecomment-2115232709, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AU7UUPIPT4JSHTQGOVNDIOTZCSXH5AVCNFSM6AAAAABFQCA3IKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMJVGIZTENZQHE. You are receiving this because you are subscribed to this thread.Message ID: @.***>

Daan-UT commented 4 months ago

I'm sorry but I see no attachment to your message @AnasEmad11

snowy199 commented 3 months ago

Did you ever figure this out?

Yes, I was recently reproducing the author's work, and I realized that this list refers to the start and end position of each video feature after concat, but this varies with the order in which the video features are concatenated

Daan-UT commented 3 months ago

How did you then find these boundaries in the concatenated list? I have another set of I3D features which are longer when concatenated than these ones, so I'd have to know the batch size

snowy199 commented 3 months ago

How did you then find these boundaries in the concatenated list? I have another set of I3D features which are longer when concatenated than these ones, so I'd have to know the batch size

I use xd's list, which is the features of each video, just load them individually, then record their lengths and splice them together, for example the first feature is [97,5,1024],the second one is [100,5,1024], and after concat it's [197,5,1024]] then the first segment is [0,96] and the second one is [97,196]

AnasEmad11 commented 3 months ago

I'm sorry but I see no attachment to your message @AnasEmad11

I have added it to ./list in the repo

JonathanFlores2503 commented 3 months ago

I'm sorry but I see no attachment to your message @AnasEmad11

I have added it to ./list in the repo

Hello, Thank you for sharing the nanlist.npy file. However, upon reviewing it, I noticed a discrepancy: the concat_UCF_V2.npy file provided contains only 779780 features [779780,10,2048], while the nanlist.npy file mentions 779951 features. Could you please provide the nanlist corresponding to the features of concat_UCF_V2.npy or advise me on how to use the shared nanlist.npy file? I appreciate your assistance.