![\"\"](\"https://imgur.com/BIRLzpb\")
I'm new in metric learning field and not have so much knowledge about this, about metric learning loss, ArcFace loss particularly. I searched the algorithm in paper and several sources, the explanation is straightforward and I understand how this comes out, but when it is implemented in code with some changes, I don't understand this so much, even paper and sources on the internet (at least all sources I found) do not mention it.
\nCan someone explain for me or give any source that has an explanation for this one. Did I miss something?
\n(Source gives first understanding of some metric learning knowledge and how it comes out for newbies (like me 🙈): https://towardsdatascience.com/the-why-and-the-how-of-deep-metric-learning-e70e16e199c0)
", "messages": [ { "id": "1698736", "postDate": "02/20/2022 16:12:36", "content": "I'm new in metric learning field and not have so much knowledge about this, about metric learning loss, ArcFace loss particularly. I searched the algorithm in paper and several sources, the explanation is straightforward and I understand how this comes out, but when it is implemented in code with some changes, I don't understand this so much, even paper and sources on the internet (at least all sources I found) do not mention it.
\nCan someone explain for me or give any source that has an explanation for this one. Did I miss something?
\n(Source gives first understanding of some metric learning knowledge and how it comes out for newbies (like me 🙈): https://towardsdatascience.com/the-why-and-the-how-of-deep-metric-learning-e70e16e199c0)
", "rawMarkdown": "I'm new in metric learning field and not have so much knowledge about this, about metric learning loss, ArcFace loss particularly. I searched the algorithm in paper and several sources, the explanation is straightforward and I understand how this comes out, but when it is implemented in code with some changes, I don't understand this so much, even paper and sources on the internet (at least all sources I found) do not mention it.\n\n\n\n\n\nCan someone explain for me or give any source that has an explanation for this one. Did I miss something?\n\n(Source gives first understanding of some metric learning knowledge and how it comes out for newbies (like me 🙈): https://towardsdatascience.com/the-why-and-the-how-of-deep-metric-learning-e70e16e199c0)", "votes": null }, { "id": "1699930", "postDate": "02/21/2022 14:30:12", "content": "I found this notebook from @slawekbiel helpful.
\n\nAlternatively, you can find many good explanations on Kaggle, Youtube videos walking you through the paper, and tutorials on the web explaining things.
\nI hope this helps.
", "rawMarkdown": "I found this notebook from @slawekbiel helpful.\n - https://www.kaggle.com/slawekbiel/arcface-explained\n\nAlternatively, you can find many good explanations on Kaggle, Youtube videos walking you through the paper, and tutorials on the web explaining things.\n\nI hope this helps.", "votes": null }, { "id": "1701205", "postDate": "02/22/2022 15:29:33", "content": "Thanks for your answer! 💯
\nI see the notebook implement according to original formula and easy for understanding, but the code implement in its repo is not.
\nBut after a few re-reading, I found out that the code still the original formula (with a few change) but with several different interpretation.
", "rawMarkdown": "Thanks for your answer! 💯\n\nI see the notebook implement according to original formula and easy for understanding, but the code implement in its repo is not.\n\nBut after a few re-reading, I found out that the code still the original formula (with a few change) but with several different interpretation.", "votes": null } ], "comments": [ { "id": 1699930, "author_name": "dschettler8845", "author_url": "", "post_date": "02/21/2022 14:30:12", "content": "I found this notebook from @slawekbiel helpful.
\n\nAlternatively, you can find many good explanations on Kaggle, Youtube videos walking you through the paper, and tutorials on the web explaining things.
\nI hope this helps.
", "votes": null, "replies": [ { "id": 1701205, "author_name": "dqhdqmcttdqx", "author_url": "", "post_date": "02/22/2022 15:29:33", "content": "Thanks for your answer! 💯
\nI see the notebook implement according to original formula and easy for understanding, but the code implement in its repo is not.
\nBut after a few re-reading, I found out that the code still the original formula (with a few change) but with several different interpretation.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1698736": "I'm new in metric learning field and not have so much knowledge about this, about metric learning loss, ArcFace loss particularly. I searched the algorithm in paper and several sources, the explanation is straightforward and I understand how this comes out, but when it is implemented in code with some changes, I don't understand this so much, even paper and sources on the internet (at least all sources I found) do not mention it.\n\n\n\n\n\nCan someone explain for me or give any source that has an explanation for this one. Did I miss something?\n\n(Source gives first understanding of some metric learning knowledge and how it comes out for newbies (like me 🙈): https://towardsdatascience.com/the-why-and-the-how-of-deep-metric-learning-e70e16e199c0)", "1699930": "I found this notebook from @slawekbiel helpful.\n - https://www.kaggle.com/slawekbiel/arcface-explained\n\nAlternatively, you can find many good explanations on Kaggle, Youtube videos walking you through the paper, and tutorials on the web explaining things.\n\nI hope this helps.", "1701205": "Thanks for your answer! 💯\n\nI see the notebook implement according to original formula and easy for understanding, but the code implement in its repo is not.\n\nBut after a few re-reading, I found out that the code still the original formula (with a few change) but with several different interpretation." }, "source": "meta" }![\"\"](\"https://imgur.com/VGnHrcp\")