For someone who's a beginner in Computer Vision, has basic knowledge of CNNs etc., what would be a good starting point for research and exploration to understand how to approach this problem?
", "messages": [ { "id": "700130", "postDate": "12/21/2019 13:57:03", "content": "For someone who's a beginner in Computer Vision, has basic knowledge of CNNs etc., what would be a good starting point for research and exploration to understand how to approach this problem?
", "rawMarkdown": "For someone who's a beginner in Computer Vision, has basic knowledge of CNNs etc., what would be a good starting point for research and exploration to understand how to approach this problem?", "votes": null }, { "id": "700137", "postDate": "12/21/2019 14:16:05", "content": "@carlossouza started some threads that you might want to read:\n- how to get started \n- he also assembled some papers on the topic you might want to read
\n\nI assembled a notebook that uses some of the models used for a recent paper on the topic:\n- https://www.kaggle.com/robikscube/faceforensics-baseline-dlib-no-internet
\n\nOr you could read their github code and paper directly:\n- https://github.com/ondyari/FaceForensics\n- https://arxiv.org/pdf/1901.08971.pdf
", "rawMarkdown": "carlossouza started some threads that you might want to read:\n- [how to get started](https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121695) \n- [he also assembled some papers on the topic you might want to read](https://www.dropbox.com/sh/ivio1l39chxalkl/AAD1_CXOMrYb8ZdAho0OqB4na?dl=0)\n\nI assembled a notebook that uses some of the models used for a recent paper on the topic:\n- https://www.kaggle.com/robikscube/faceforensics-baseline-dlib-no-internet\n\nOr you could read their github code and paper directly:\n- https://github.com/ondyari/FaceForensics\n- https://arxiv.org/pdf/1901.08971.pdf", "votes": null }, { "id": "700142", "postDate": "12/21/2019 14:19:40", "content": "Thanks!
", "rawMarkdown": "Thanks!", "votes": null }, { "id": "707222", "postDate": "12/31/2019 12:41:14", "content": "Thanks for the information.
", "rawMarkdown": "Thanks for the information.", "votes": null }, { "id": "707234", "postDate": "12/31/2019 13:02:53", "content": "A very simple approach is to find the face in one or more frames of every video and save that face region to a PNG file in a \"fake\" or \"real\" folder (based on the label for that video). Then train a binary image classifier that can tell fake faces apart from real faces.
", "rawMarkdown": "A very simple approach is to find the face in one or more frames of every video and save that face region to a PNG file in a \"fake\" or \"real\" folder (based on the label for that video). Then train a binary image classifier that can tell fake faces apart from real faces.", "votes": null } ], "comments": [ { "id": 700137, "author_name": "robikscube", "author_url": "", "post_date": "12/21/2019 14:16:05", "content": "@carlossouza started some threads that you might want to read:\n- how to get started \n- he also assembled some papers on the topic you might want to read
\n\nI assembled a notebook that uses some of the models used for a recent paper on the topic:\n- https://www.kaggle.com/robikscube/faceforensics-baseline-dlib-no-internet
\n\nOr you could read their github code and paper directly:\n- https://github.com/ondyari/FaceForensics\n- https://arxiv.org/pdf/1901.08971.pdf
", "votes": null, "replies": [ { "id": 700142, "author_name": "shrutimechlearn", "author_url": "", "post_date": "12/21/2019 14:19:40", "content": "Thanks!
", "votes": null, "replies": [] }, { "id": 707222, "author_name": "pidugusundeep", "author_url": "", "post_date": "12/31/2019 12:41:14", "content": "Thanks for the information.
", "votes": null, "replies": [] } ] }, { "id": 707234, "author_name": "humananalog", "author_url": "", "post_date": "12/31/2019 13:02:53", "content": "A very simple approach is to find the face in one or more frames of every video and save that face region to a PNG file in a \"fake\" or \"real\" folder (based on the label for that video). Then train a binary image classifier that can tell fake faces apart from real faces.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "700130": "For someone who's a beginner in Computer Vision, has basic knowledge of CNNs etc., what would be a good starting point for research and exploration to understand how to approach this problem?", "700137": "carlossouza started some threads that you might want to read:\n- [how to get started](https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121695) \n- [he also assembled some papers on the topic you might want to read](https://www.dropbox.com/sh/ivio1l39chxalkl/AAD1_CXOMrYb8ZdAho0OqB4na?dl=0)\n\nI assembled a notebook that uses some of the models used for a recent paper on the topic:\n- https://www.kaggle.com/robikscube/faceforensics-baseline-dlib-no-internet\n\nOr you could read their github code and paper directly:\n- https://github.com/ondyari/FaceForensics\n- https://arxiv.org/pdf/1901.08971.pdf", "700142": "Thanks!", "707222": "Thanks for the information.", "707234": "A very simple approach is to find the face in one or more frames of every video and save that face region to a PNG file in a \"fake\" or \"real\" folder (based on the label for that video). Then train a binary image classifier that can tell fake faces apart from real faces." }, "source": "meta" }