![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F10356799%2Ffb191cdb05c65978b311daf8fdd79cf1%2FScreenshot%202025-03-12%20at%206.51.44.png?generation=1741758747637983&alt=media\")
Datasets with 224x224 png melspectrograms for 5sec and 10sec chunks. Could be useful for training.
\n\n
Datasets with 224x224 png melspectrograms for 5sec and 10sec chunks. Could be useful for training.
\n\n
He's not the hero we deserve, but the hero we need
", "rawMarkdown": "He's not the hero we deserve, but the hero we need", "votes": 2 }, { "id": 3171444, "postDate": "2025-04-05T17:25:06.363Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 3148670, "postDate": "2025-03-13T12:06:12.580Z", "content": "Hi @samvelkoch Thanks for the extensive work you did. I have a question as you mentioned that 1 file is divided into multiple chunks like if audio file is 45 sec long you divided into 9, 5 seconds chunks. This will be a lot of data to train and currently I am solely depend on kaggle GPU's do you have better strategy to chunk out the data or do we need to proceed like this only.
\nThanks again
There are multiple approaches to deal with data in this competition. Prepared datasets is not the only way for sure. It just can be useful sometime to save computing time. But I would definitely suggest you to research for best practices form last years CLEFs.
", "rawMarkdown": "There are multiple approaches to deal with data in this competition. Prepared datasets is not the only way for sure. It just can be useful sometime to save computing time. But I would definitely suggest you to research for best practices form last years CLEFs. " } ] } ], "comments": [ { "id": 3147554, "author_name": "eikyou", "author_url": "", "post_date": "2025-03-12T06:03:39.893000", "content": "He's not the hero we deserve, but the hero we need
", "votes": 2, "replies": [] }, { "id": 3171444, "author_name": "", "author_url": "", "post_date": "2025-04-05T17:25:06.363000", "content": "", "votes": 1, "replies": [] }, { "id": 3148670, "author_name": "", "author_url": "", "post_date": "2025-03-13T12:06:12.580000", "content": "Hi @samvelkoch Thanks for the extensive work you did. I have a question as you mentioned that 1 file is divided into multiple chunks like if audio file is 45 sec long you divided into 9, 5 seconds chunks. This will be a lot of data to train and currently I am solely depend on kaggle GPU's do you have better strategy to chunk out the data or do we need to proceed like this only.
\nThanks again
There are multiple approaches to deal with data in this competition. Prepared datasets is not the only way for sure. It just can be useful sometime to save computing time. But I would definitely suggest you to research for best practices form last years CLEFs.
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "3147551": "Datasets with 224x224 png melspectrograms for 5sec and 10sec chunks. Could be useful for training. \n\n- [BirdCLEF+ 2025, MelSPECs (5 sec)](https://www.kaggle.com/datasets/samvelkoch/birdclef-2025-melspecs-5-sec/)\n- [BirdCLEF+ 2025, MelSPECs (10 sec)](https://www.kaggle.com/datasets/samvelkoch/birdclef-2025-melspecs-10-sec)\n\n\n", "3147554": "He's not the hero we deserve, but the hero we need", "3171444": "", "3148670": "Hi @samvelkoch Thanks for the extensive work you did. I have a question as you mentioned that 1 file is divided into multiple chunks like if audio file is 45 sec long you divided into 9, 5 seconds chunks. This will be a lot of data to train and currently I am solely depend on kaggle GPU's do you have better strategy to chunk out the data or do we need to proceed like this only.\nThanks again" } }