search

dk-liang / CLTR

[ECCV 2022] An End-to-End Transformer Model for Crowd Localization
MIT License
87 stars 13 forks source link

KMO based Matcher #5

Closed ls-ash closed 1 year ago

ls-ash commented 1 year ago

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?

ds2268 commented 1 year ago

Yes, I can also only see the basic L1 implementation of the matcher.

ds2268 commented 1 year ago

@dk-liang: Can you also provide a KMO-based mather in order to reproduce the results in new publications?

seungyonglee0802 commented 1 year ago

@ds2268 Did you get KMO-based matcher? I can't find it in this repository...

congyi-lcy commented 1 year ago

I met the same problem. Did you solve it?

Yes, I can also only see the basic L1 implementation of the matcher.

seungyonglee0802 commented 1 year ago

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 🤣

congyi-lcy commented 1 year ago

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: @.***>

seungyonglee0802 commented 1 year ago

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 :(

congyi-lcy commented 1 year ago

Well. thank you for replying and communicating together ---- Replied Message ---- | From | @.> | | Date | 07/10/2023 22:53 | | To | dk-liang/CLTR @.> | | Cc | @.>, Comment @.> | | Subject | Re: [dk-liang/CLTR] KMO based Matcher (Issue #5) |

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 :(

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>