I have published a notebook regarding how we can improve our score using 2 class filter from 14 class prediction.
\nIt improves the LB score from 0.154 to 0.201.
I have published a notebook regarding how we can improve our score using 2 class filter from 14 class prediction.
\nIt improves the LB score from 0.154 to 0.201.
Can i ask you about how did you create COCO format dataset ?
", "rawMarkdown": "Can i ask you about how did you create COCO format dataset ?\n", "votes": 1, "replies": [ { "id": 1154122, "postDate": "2021-01-15T12:22:57.830Z", "content": "Who said I did ?😐
", "rawMarkdown": "Who said I did ?😐" }, { "id": 1154127, "postDate": "2021-01-15T12:27:59.877Z", "content": "oh. i thinked that. not ?
", "rawMarkdown": "oh. i thinked that. not ?" }, { "id": 1154129, "postDate": "2021-01-15T12:29:06.370Z", "content": "https://www.kaggle.com/awsaf49/vinbigdata-256-image-dataset . It's your dataset
", "rawMarkdown": "https://www.kaggle.com/awsaf49/vinbigdata-256-image-dataset . It's your dataset" }, { "id": 1154177, "postDate": "2021-01-15T13:18:28.180Z", "content": "Yes, but I didn't convert the annotations so it's in the original form.
", "rawMarkdown": "Yes, but I didn't convert the annotations so it's in the original form." } ] }, { "id": 1148268, "postDate": "2021-01-11T03:28:46.610Z", "content": "Thanks for sharing, 2-class classification also is the first stage in my pipeline! But I am afraid of overfitting since the public test set could contain many 'No finding' cases and yet the public test just comprises 10% of the full data set.
", "rawMarkdown": "Thanks for sharing, 2-class classification also is the first stage in my pipeline! But I am afraid of overfitting since the public test set could contain many 'No finding' cases and yet the public test just comprises 10% of the full data set. ", "votes": 1 }, { "id": 1147553, "postDate": "2021-01-10T15:17:04.403Z", "content": "Thanks for sharing, it's very helpful for newbie or someone who wants to join this competition. I don't understand why it's downvoted…
", "rawMarkdown": "Thanks for sharing, it's very helpful for newbie or someone who wants to join this competition. I don't understand why it's downvoted...", "votes": 1, "replies": [ { "id": 1147575, "postDate": "2021-01-10T15:34:59.183Z", "content": "I guess I have somehow spoiled his idea by sharing 😬
", "rawMarkdown": "I guess I have somehow spoiled his idea by sharing 😬" } ] }, { "id": 1151906, "postDate": "2021-01-13T16:30:53.430Z", "content": "Thanks for sharing! @awsaf49
", "rawMarkdown": "Thanks for sharing! @awsaf49 ", "votes": 1 }, { "id": 1147814, "postDate": "2021-01-10T18:02:59.383Z", "content": "Nice work! Thanks for that!
", "rawMarkdown": "Nice work! Thanks for that!", "votes": 1 } ], "comments": [ { "id": 1154116, "author_name": "Manh Lab", "author_url": "", "post_date": "2021-01-15T12:18:33.747000", "content": "Can i ask you about how did you create COCO format dataset ?
", "votes": 1, "replies": [ { "id": 1154122, "author_name": "Awsaf", "author_url": "", "post_date": "2021-01-15T12:22:57.830000", "content": "Who said I did ?😐
", "votes": 0, "replies": [] }, { "id": 1154127, "author_name": "Manh Lab", "author_url": "", "post_date": "2021-01-15T12:27:59.877000", "content": "oh. i thinked that. not ?
", "votes": 0, "replies": [] }, { "id": 1154129, "author_name": "Manh Lab", "author_url": "", "post_date": "2021-01-15T12:29:06.370000", "content": "https://www.kaggle.com/awsaf49/vinbigdata-256-image-dataset . It's your dataset
", "votes": 0, "replies": [] }, { "id": 1154177, "author_name": "Awsaf", "author_url": "", "post_date": "2021-01-15T13:18:28.180000", "content": "Yes, but I didn't convert the annotations so it's in the original form.
", "votes": 0, "replies": [] } ] }, { "id": 1148268, "author_name": "The fearless", "author_url": "", "post_date": "2021-01-11T03:28:46.610000", "content": "Thanks for sharing, 2-class classification also is the first stage in my pipeline! But I am afraid of overfitting since the public test set could contain many 'No finding' cases and yet the public test just comprises 10% of the full data set.
", "votes": 1, "replies": [] }, { "id": 1147553, "author_name": "Camaro", "author_url": "", "post_date": "2021-01-10T15:17:04.403000", "content": "Thanks for sharing, it's very helpful for newbie or someone who wants to join this competition. I don't understand why it's downvoted…
", "votes": 1, "replies": [ { "id": 1147575, "author_name": "Awsaf", "author_url": "", "post_date": "2021-01-10T15:34:59.183000", "content": "I guess I have somehow spoiled his idea by sharing 😬
", "votes": 0, "replies": [] } ] }, { "id": 1151906, "author_name": "Satyajit Pattnaik", "author_url": "", "post_date": "2021-01-13T16:30:53.430000", "content": "Thanks for sharing! @awsaf49
", "votes": 1, "replies": [] }, { "id": 1147814, "author_name": "Manh Lab", "author_url": "", "post_date": "2021-01-10T18:02:59.383000", "content": "Nice work! Thanks for that!
", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "1147433": "I have published a [notebook](https://www.kaggle.com/awsaf49/vinbigdata-2-class-filter) regarding how we can improve our score using `2 class filter` from `14 class` prediction.\nIt improves the **LB** score from `0.154` to `0.201`.\n* [VinBigData 🌟2 Class Filter🌟](https://www.kaggle.com/awsaf49/vinbigdata-2-class-filter) [LB: 0.201]\n* [14 class train](https://www.kaggle.com/awsaf49/vinbigdata-cxr-ad-yolov5-14-class-train)\n* [14 class infer](https://www.kaggle.com/awsaf49/vinbigdata-cxr-ad-yolov5-14-class-infer) [LB: 0.154]", "1154116": "Can i ask you about how did you create COCO format dataset ?\n", "1148268": "Thanks for sharing, 2-class classification also is the first stage in my pipeline! But I am afraid of overfitting since the public test set could contain many 'No finding' cases and yet the public test just comprises 10% of the full data set. ", "1147553": "Thanks for sharing, it's very helpful for newbie or someone who wants to join this competition. I don't understand why it's downvoted...", "1151906": "Thanks for sharing! @awsaf49 ", "1147814": "Nice work! Thanks for that!" } }