{ "id": 412806, "title": "Recent promising CVPR2023 papers on image matching.", "url": "/competitions/image-matching-challenge-2023/discussion/412806", "author_name": "old-ufo", "post_date": "2023-05-25T10:34:07.789000", "votes": 42, "comment_count": 4, "views": 0, "content": "

I'd like to share some recent papers on image matching with repos, which might or might not be helpful for this competition.

\n", "messages": [ { "id": 2273678, "postDate": "2023-05-25T10:34:07.790Z", "content": "

I'd like to share some recent papers on image matching with repos, which might or might not be helpful for this competition.

\n", "rawMarkdown": "I'd like to share some recent papers on image matching with repos, which might or might not be helpful for this competition. \n\n- SiLK. https://arxiv.org/abs/2304.06194, https://github.com/facebookresearch/silk They claim to be better than DiSK and even SuperGlue.\n- DAC https://github.com/javrtg/DAC https://arxiv.org/abs/2305.12250 . Post-training estimation of keypoints reliability such as SuperPoint and DISK.\n- DKM. Much denser version of LoFTR with uncertainty estimation, good performance in IMC-2022 https://arxiv.org/abs/2202.00667 https://github.com/Parskatt/DKM \n- IMP https://arxiv.org/pdf/2304.14837.pdf https://github.com/feixue94/imp-release Bad license though.\n- AdaMatcher https://github.com/TencentYoutuResearch/AdaMatcher https://arxiv.org/abs/2207.08427\n- MTLDesc https://github.com/vignywang/MTLDesc https://arxiv.org/abs/2203.07003\n- Two-view Geometry Scoring Without Correspondences https://openaccess.thecvf.com/content/CVPR2023/papers/Barroso-Laguna_Two-View_Geometry_Scoring_Without_Correspondences_CVPR_2023_paper.pdf https://github.com/nianticlabs/scoring-without-correspondences \n- NCMNet - OANet/CLNet improvements https://openaccess.thecvf.com/content/CVPR2023/papers/Liu_Progressive_Neighbor_Consistency_Mining_for_Correspondence_Pruning_CVPR_2023_paper.pdf https://github.com/xinliu29/NCMNet ", "votes": 41 }, { "id": 2273724, "postDate": "2023-05-25T11:12:02.710Z", "content": "

@oldufo Hey Pal thanks for searching and sharing these papers with us.

", "rawMarkdown": "@oldufo Hey Pal thanks for searching and sharing these papers with us." }, { "id": 2280549, "postDate": "2023-05-30T07:13:45.977Z", "content": "

I'm sorry, I cannot submit using SiLK.

\n

Your Notebook uses non-versioned datasets [/datasets/piezzo/silk-kp] (see Dataset Settings).

", "rawMarkdown": "I'm sorry, I cannot submit using SiLK. \n\nYour Notebook uses non-versioned datasets [/datasets/piezzo/silk-kp] (see Dataset Settings)." }, { "id": 2278355, "postDate": "2023-05-28T15:53:49.747Z", "content": "

nice work,bro!

", "rawMarkdown": "nice work,bro!" }, { "id": 2274601, "postDate": "2023-05-26T06:09:59.073Z", "rawMarkdown": "", "votes": -1, "isDeleted": true } ], "comments": [ { "id": 2273724, "author_name": "Muhammad Bilal Hussain", "author_url": "", "post_date": "2023-05-25T11:12:02.710000", "content": "

@oldufo Hey Pal thanks for searching and sharing these papers with us.

", "votes": 0, "replies": [] }, { "id": 2280549, "author_name": "Yanbing Zhang", "author_url": "", "post_date": "2023-05-30T07:13:45.977000", "content": "

I'm sorry, I cannot submit using SiLK.

\n

Your Notebook uses non-versioned datasets [/datasets/piezzo/silk-kp] (see Dataset Settings).

", "votes": 0, "replies": [] }, { "id": 2278355, "author_name": "xiayu", "author_url": "", "post_date": "2023-05-28T15:53:49.747000", "content": "

nice work,bro!

", "votes": 0, "replies": [] }, { "id": 2274601, "author_name": "", "author_url": "", "post_date": "2023-05-26T06:09:59.073000", "content": "", "votes": -1, "replies": [] } ], "raw_markdown_by_id": { "2273678": "I'd like to share some recent papers on image matching with repos, which might or might not be helpful for this competition. \n\n- SiLK. https://arxiv.org/abs/2304.06194, https://github.com/facebookresearch/silk They claim to be better than DiSK and even SuperGlue.\n- DAC https://github.com/javrtg/DAC https://arxiv.org/abs/2305.12250 . Post-training estimation of keypoints reliability such as SuperPoint and DISK.\n- DKM. Much denser version of LoFTR with uncertainty estimation, good performance in IMC-2022 https://arxiv.org/abs/2202.00667 https://github.com/Parskatt/DKM \n- IMP https://arxiv.org/pdf/2304.14837.pdf https://github.com/feixue94/imp-release Bad license though.\n- AdaMatcher https://github.com/TencentYoutuResearch/AdaMatcher https://arxiv.org/abs/2207.08427\n- MTLDesc https://github.com/vignywang/MTLDesc https://arxiv.org/abs/2203.07003\n- Two-view Geometry Scoring Without Correspondences https://openaccess.thecvf.com/content/CVPR2023/papers/Barroso-Laguna_Two-View_Geometry_Scoring_Without_Correspondences_CVPR_2023_paper.pdf https://github.com/nianticlabs/scoring-without-correspondences \n- NCMNet - OANet/CLNet improvements https://openaccess.thecvf.com/content/CVPR2023/papers/Liu_Progressive_Neighbor_Consistency_Mining_for_Correspondence_Pruning_CVPR_2023_paper.pdf https://github.com/xinliu29/NCMNet ", "2273724": "@oldufo Hey Pal thanks for searching and sharing these papers with us.", "2280549": "I'm sorry, I cannot submit using SiLK. \n\nYour Notebook uses non-versioned datasets [/datasets/piezzo/silk-kp] (see Dataset Settings).", "2278355": "nice work,bro!", "2274601": "" } }