![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F933480%2F3bdfefb1dda818e9c990ff6a5d95c44d%2Fmvp.jpg?generation=1601485541571324&alt=media\")
Does anyone know if there is a PyTorch implementation of DELG floating around anywhere? I don't recall where but I think I recall perhaps @aerdem4 or maybe some other GM mentioning they had one? It would be really useful if anyone could share a copy, even if it's not official.
\nšš¾
", "messages": [ { "id": "1032289", "postDate": "09/30/2020 05:27:19", "content": "Does anyone know if there is a PyTorch implementation of DELG floating around anywhere? I don't recall where but I think I recall perhaps @aerdem4 or maybe some other GM mentioning they had one? It would be really useful if anyone could share a copy, even if it's not official.
\nšš¾
", "rawMarkdown": "Does anyone know if there is a PyTorch implementation of DELG floating around anywhere? I don't recall where but I think I recall perhaps @aerdem4 or maybe some other GM mentioning they had one? It would be really useful if anyone could share a copy, even if it's not official.\n\nšš¾", "votes": null }, { "id": "1032535", "postDate": "09/30/2020 09:14:36", "content": "I am also interested in this. Maybe even setup a git repo. I worked on porting the weights and have some snippets but did not finally check if output is exactly the same.
", "rawMarkdown": "I am also interested in this. Maybe even setup a git repo. I worked on porting the weights and have some snippets but did not finally check if output is exactly the same.", "votes": null }, { "id": "1032852", "postDate": "09/30/2020 13:31:14", "content": "I am interest in this as well.
\nCongratulations @authman.
", "rawMarkdown": "I am interest in this as well.\n\nCongratulations @authman.", "votes": null }, { "id": "1032858", "postDate": "09/30/2020 13:39:13", "content": "Hey, did you mean this discussion?:
\nhttps://www.kaggle.com/c/landmark-retrieval-2020/discussion/176151
Thank you. My team, and in particular @madz2000, really did all the heavy lifting on this one.
", "rawMarkdown": "Thank you. My team, and in particular @madz2000, really did all the heavy lifting on this one.", "votes": null }, { "id": "1032899", "postDate": "09/30/2020 14:01:27", "content": "Ah! Yes, this is where I read it..
", "rawMarkdown": "Ah! Yes, this is where I read it..", "votes": null }, { "id": "1032927", "postDate": "09/30/2020 14:15:03", "content": "I think there is a repo I stumbled upon very early into the competition(just going through some stuff) , just try searching , it wasn't that difficult to find.
", "rawMarkdown": "I think there is a repo I stumbled upon very early into the competition(just going through some stuff) , just try searching , it wasn't that difficult to find.", "votes": null }, { "id": "1032962", "postDate": "09/30/2020 14:56:30", "content": "Thanks @authman. I learnt a lot of new things in this competition from you and @tanulsingh077. Looking forward to further collaborations with you all :)
", "rawMarkdown": "Thanks @authman. I learnt a lot of new things in this competition from you and @tanulsingh077. Looking forward to further collaborations with you all :)", "votes": null }, { "id": "1033019", "postDate": "09/30/2020 15:44:35", "content": "Please update here if you found one.
", "rawMarkdown": "Please update here if you found one.", "votes": null }, { "id": "1033056", "postDate": "09/30/2020 16:19:58", "content": "@authman, I did not realize you competed with a team. Congrats @madz2000, @tanulsingh077, @phoenix9032, and @kshitijmohan.
", "rawMarkdown": "authman, I did not realize you competed with a team. Congrats @madz2000, @tanulsingh077, @phoenix9032, and @kshitijmohan.", "votes": null }, { "id": "1033072", "postDate": "09/30/2020 16:33:50", "content": "I'm writing our team's solution, then we're gonna release the code + model checkpoints. Some nice stuffs such as DELG, SuperPoint's pytorch port, Seesaw loss will be there š
", "rawMarkdown": "I'm writing our team's solution, then we're gonna release the code + model checkpoints. Some nice stuffs such as DELG, SuperPoint's pytorch port, Seesaw loss will be there š", "votes": null }, { "id": "1033109", "postDate": "09/30/2020 17:05:21", "content": "
Thank you!
", "rawMarkdown": "Thank you!", "votes": null }, { "id": "1033145", "postDate": "09/30/2020 17:39:02", "content": "Thanks @sheriytm it was mostly just tweaking the ransac params and @madz2000 was our go to man here . Looking forward to see the DELG model training code and winning solutions to learn
", "rawMarkdown": "Thanks @sheriytm it was mostly just tweaking the ransac params and @madz2000 was our go to man here . Looking forward to see the DELG model training code and winning solutions to learn", "votes": null }, { "id": "1033174", "postDate": "09/30/2020 18:11:55", "content": "@tanulsingh077, that is a very good result by tweaks.
\nMy 1st submission was done by my 4yr old nephew by mistake. I joined the contest but had no intention on making a submission. I ran a couple of experiments, then let my nephew play games on my tablet without closing all my windows and he somehow managed to submit one of the kernels. So I had to make a few more submissions by tweaking like you did and closed shop. I just did not have the time to participate here.
", "rawMarkdown": "tanulsingh077, that is a very good result by tweaks. \n\nMy 1st submission was done by my 4yr old nephew by mistake. I joined the contest but had no intention on making a submission. I ran a couple of experiments, then let my nephew play games on my tablet without closing all my windows and he somehow managed to submit one of the kernels. So I had to make a few more submissions by tweaking like you did and closed shop. I just did not have the time to participate here.", "votes": null } ], "comments": [ { "id": 1032535, "author_name": "christofhenkel", "author_url": "", "post_date": "09/30/2020 09:14:36", "content": "I am also interested in this. Maybe even setup a git repo. I worked on porting the weights and have some snippets but did not finally check if output is exactly the same.
", "votes": null, "replies": [] }, { "id": 1032852, "author_name": "sheriytm", "author_url": "", "post_date": "09/30/2020 13:31:14", "content": "I am interest in this as well.
\nCongratulations @authman.
", "votes": null, "replies": [ { "id": 1032896, "author_name": "authman", "author_url": "", "post_date": "09/30/2020 14:00:32", "content": "Thank you. My team, and in particular @madz2000, really did all the heavy lifting on this one.
", "votes": null, "replies": [] }, { "id": 1032962, "author_name": "madz2000", "author_url": "", "post_date": "09/30/2020 14:56:30", "content": "Thanks @authman. I learnt a lot of new things in this competition from you and @tanulsingh077. Looking forward to further collaborations with you all :)
", "votes": null, "replies": [] }, { "id": 1033056, "author_name": "sheriytm", "author_url": "", "post_date": "09/30/2020 16:19:58", "content": "@authman, I did not realize you competed with a team. Congrats @madz2000, @tanulsingh077, @phoenix9032, and @kshitijmohan.
", "votes": null, "replies": [] }, { "id": 1033113, "author_name": "madz2000", "author_url": "", "post_date": "09/30/2020 17:09:53", "content": "Thank you!
", "votes": null, "replies": [] }, { "id": 1033145, "author_name": "tanulsingh077", "author_url": "", "post_date": "09/30/2020 17:39:02", "content": "Thanks @sheriytm it was mostly just tweaking the ransac params and @madz2000 was our go to man here . Looking forward to see the DELG model training code and winning solutions to learn
", "votes": null, "replies": [] }, { "id": 1033174, "author_name": "sheriytm", "author_url": "", "post_date": "09/30/2020 18:11:55", "content": "@tanulsingh077, that is a very good result by tweaks.
\nMy 1st submission was done by my 4yr old nephew by mistake. I joined the contest but had no intention on making a submission. I ran a couple of experiments, then let my nephew play games on my tablet without closing all my windows and he somehow managed to submit one of the kernels. So I had to make a few more submissions by tweaking like you did and closed shop. I just did not have the time to participate here.
", "votes": null, "replies": [] } ] }, { "id": 1032858, "author_name": "behlarjun", "author_url": "", "post_date": "09/30/2020 13:39:13", "content": "Hey, did you mean this discussion?:
\nhttps://www.kaggle.com/c/landmark-retrieval-2020/discussion/176151
Ah! Yes, this is where I read it..
", "votes": null, "replies": [] }, { "id": 1032927, "author_name": "behlarjun", "author_url": "", "post_date": "09/30/2020 14:15:03", "content": "I think there is a repo I stumbled upon very early into the competition(just going through some stuff) , just try searching , it wasn't that difficult to find.
", "votes": null, "replies": [] } ] }, { "id": 1033019, "author_name": "rahim3", "author_url": "", "post_date": "09/30/2020 15:44:35", "content": "Please update here if you found one.
", "votes": null, "replies": [] }, { "id": 1033072, "author_name": "andy2709", "author_url": "", "post_date": "09/30/2020 16:33:50", "content": "I'm writing our team's solution, then we're gonna release the code + model checkpoints. Some nice stuffs such as DELG, SuperPoint's pytorch port, Seesaw loss will be there š
", "votes": null, "replies": [ { "id": 1033109, "author_name": "authman", "author_url": "", "post_date": "09/30/2020 17:05:21", "content": "