KMO based Matcher
ls-ash opened this issue · comments
After reading your article and code, I am interested in the part of Matcher, but I do not find KMO based Matcher in the code. Is it in the module or something else?
Yes, I can also only see the basic L1 implementation of the matcher.
@dk-liang: Can you also provide a KMO-based mather in order to reproduce the results in new publications?
@ds2268 Did you get KMO-based matcher? I can't find it in this repository...
I met the same problem. Did you solve it?
Yes, I can also only see the basic L1 implementation of the matcher.
I met the same problem. Did you solve it?
Yes, I can also only see the basic L1 implementation of the matcher.
No.. not yet 🤣
All right... That is，it feels like this paper just moved DETR to the crowd counting task，without any other greater improvements😵💫 ---- Replied Message ---- | From | @.> | | Date | 07/10/2023 22:41 | | To | dk-liang/CLTR @.> | | Cc | @.>, Comment @.> | | Subject | Re: [dk-liang/CLTR] KMO based Matcher (Issue #5) | I met the same problem. Did you solve it? Yes, I can also only see the basic L1 implementation of the matcher. No.. not yet 🤣 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>
I agree, this paper only use conditional DETR without bounding boxes and class information. I think a main novelty is the KMO matcher, but they didn't open this as a open source :(