I am feeling secondhand embarrassment for our competition hosts... I think they forgot about how should work with hidden test in code competition. It is normal (not).
\n\nSo, I have created custom dataset for making check phase. It has some hardcode (with one line), but it will help you avoid many bugs without button \"submission\".
\n\n\n\nAlso I added my custom method for super fast reading audio (read all train without exceptions). It uses torchaudio and librosa (if need).
\n\nSubmission with reading hidden test audio has time ~3-4 min (without using model).
\n\nWelcome!
\n\nP.S. I noticed some downvotes If you think that I am too rude with competition hosts or kaggle team, then you should read this topic and all messages. It is my open position. I love kaggle community (all users and hardworking competitors), but I don't like when organizers ignore and don't make concessions against really good arguments. But I have to say thanks for interesting competition with audio!
", "messages": [ { "id": "893102", "postDate": "06/19/2020 11:46:31", "content": "I am feeling secondhand embarrassment for our competition hosts... I think they forgot about how should work with hidden test in code competition. It is normal (not).
\n\nSo, I have created custom dataset for making check phase. It has some hardcode (with one line), but it will help you avoid many bugs without button \"submission\".
\n\n\n\nAlso I added my custom method for super fast reading audio (read all train without exceptions). It uses torchaudio and librosa (if need).
\n\nSubmission with reading hidden test audio has time ~3-4 min (without using model).
\n\nWelcome!
\n\nP.S. I noticed some downvotes If you think that I am too rude with competition hosts or kaggle team, then you should read this topic and all messages. It is my open position. I love kaggle community (all users and hardworking competitors), but I don't like when organizers ignore and don't make concessions against really good arguments. But I have to say thanks for interesting competition with audio!
", "rawMarkdown": "I am feeling secondhand embarrassment for our competition hosts... I think they [forgot about how should work with hidden test in code competition](https://www.kaggle.com/c/birdsong-recognition/discussion/158987#892424). It is normal (not).\n\nSo, I have created [custom dataset](https://www.kaggle.com/shonenkov/birdcall-check) for making check phase. It has some hardcode (with one line), but it will help you avoid many bugs without button \"submission\".\n\n- [[Sample Submission] Using Custom Check](https://www.kaggle.com/shonenkov/sample-submission-using-custom-check)\n\nAlso I added my custom method for super fast reading audio (read all train without exceptions). It uses torchaudio and librosa (if need). \n\nSubmission with reading hidden test audio has time ~3-4 min (without using model). \n\nWelcome!\n\n\nP.S. I noticed some downvotes If you think that I am too rude with competition hosts or kaggle team, then you should read [this topic and all messages](https://www.kaggle.com/c/deepfake-detection-challenge/discussion/157983#886135). It is my open position. I love kaggle community (all users and hardworking competitors), but I don't like when organizers ignore and don't make concessions against really good arguments. But I have to say thanks for interesting competition with audio!", "votes": null }, { "id": "902641", "postDate": "06/26/2020 09:24:33", "content": "Thank you for the effort, that's really helpful.
", "rawMarkdown": "Thank you for the effort, that's really helpful.", "votes": null }, { "id": "912137", "postDate": "07/02/2020 08:53:49", "content": "Thanks a lot!
", "rawMarkdown": "Thanks a lot!", "votes": null }, { "id": "990363", "postDate": "08/29/2020 14:58:20", "content": "Thanks!
", "rawMarkdown": "Thanks!", "votes": null }, { "id": "993330", "postDate": "08/31/2020 21:00:16", "content": "thanks for this! helped me test in non-blind manner)
", "rawMarkdown": "thanks for this! helped me test in non-blind manner)", "votes": null }, { "id": "995356", "postDate": "09/02/2020 11:31:31", "content": "@shonenkov I follow most of your contributions and learning, still learning a lot. NLP albumentations is my favourite one for your work.
", "rawMarkdown": "shonenkov I follow most of your contributions and learning, still learning a lot. NLP albumentations is my favourite one for your work.", "votes": null } ], "comments": [ { "id": 993330, "author_name": "kruntuid", "author_url": "", "post_date": "08/31/2020 21:00:16", "content": "thanks for this! helped me test in non-blind manner)
", "votes": null, "replies": [] }, { "id": 995356, "author_name": "validmodel", "author_url": "", "post_date": "09/02/2020 11:31:31", "content": "@shonenkov I follow most of your contributions and learning, still learning a lot. NLP albumentations is my favourite one for your work.
", "votes": null, "replies": [] }, { "id": 902641, "author_name": "CVxTz", "author_url": "", "post_date": "06/26/2020 09:24:33", "content": "Thank you for the effort, that's really helpful.
", "votes": null, "replies": [] }, { "id": 912137, "author_name": "nejel34", "author_url": "", "post_date": "07/02/2020 08:53:49", "content": "Thanks a lot!
", "votes": null, "replies": [] }, { "id": 990363, "author_name": "tridibdas", "author_url": "", "post_date": "08/29/2020 14:58:20", "content": "Thanks!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "893102": "I am feeling secondhand embarrassment for our competition hosts... I think they [forgot about how should work with hidden test in code competition](https://www.kaggle.com/c/birdsong-recognition/discussion/158987#892424). It is normal (not).\n\nSo, I have created [custom dataset](https://www.kaggle.com/shonenkov/birdcall-check) for making check phase. It has some hardcode (with one line), but it will help you avoid many bugs without button \"submission\".\n\n- [[Sample Submission] Using Custom Check](https://www.kaggle.com/shonenkov/sample-submission-using-custom-check)\n\nAlso I added my custom method for super fast reading audio (read all train without exceptions). It uses torchaudio and librosa (if need). \n\nSubmission with reading hidden test audio has time ~3-4 min (without using model). \n\nWelcome!\n\n\nP.S. I noticed some downvotes If you think that I am too rude with competition hosts or kaggle team, then you should read [this topic and all messages](https://www.kaggle.com/c/deepfake-detection-challenge/discussion/157983#886135). It is my open position. I love kaggle community (all users and hardworking competitors), but I don't like when organizers ignore and don't make concessions against really good arguments. But I have to say thanks for interesting competition with audio!", "902641": "Thank you for the effort, that's really helpful.", "912137": "Thanks a lot!", "990363": "Thanks!", "993330": "thanks for this! helped me test in non-blind manner)", "995356": "shonenkov I follow most of your contributions and learning, still learning a lot. NLP albumentations is my favourite one for your work." }, "source": "meta" }