{ "id": 105305, "title": "Must-see Kernels for 'APTOS 2019 Blindness Detection'", "url": "/competitions/aptos2019-blindness-detection/discussion/105305", "author_name": "datartist", "post_date": "2019-08-22T08:20:43.376000", "votes": 46, "comment_count": 12, "views": 0, "content": "

[1] All in one\n- Intro APTOS Diabetic Retinopathy (EDA & Starter)\n14000+ views, 350+ upvotes\n- APTOS 2019: DenseNet Keras Starter\n15000+ views, 270+ upvotes\n- APTOS [updated] Albumentation meets Grad-CAM\n2600+ views, 80+ upvotes\n- Basic EDA and baseline pytorch model\n2900+ views, 70+ upvotes\n- Keras+ResNet+Test Time Augmentation\n6000+ views, 50+ upvotes\n- APTOS Blindness Detection - EDA and Keras ResNet50\n2200+ views, 50+ upvotes

\n\n

[2] EDA\n- Image shape distribution - previous and present\n1200+ views, 30+ upvotes\n- APTOS2019: Blindness Detection EDA\n600+ views, 20+ upvotes\n- Basic EDA - train test image distribution check\n600+ views, 10+ upvotes

\n\n

[3] Baseline\n- [APTOS] resnet50 baseline\n10000 views, 140+ upvotes\n- Eye -- EfficientNet Pytorch[LB 0.777] \n5500+ views, 100+ upvotes\n- EfficientNetB5 with Keras (APTOS 2019)\n4300+ views, 90+ upvotes\n- APTOS 2019: Keras Baseline\n5100+ views, 90+ upvotes\n- EfficientNetB4 FastAI - Blindness Detection\n4800+ views, 70+ upvotes

\n\n

[4] Preprocessing\n- APTOS [UpdatedV14] Preprocessing- Ben's & Cropping\n21000+ views, 560+ upvotes\n- Pre-processing train and test images\n4000+ views, 120+ upvotes

\n\n

[5] Optimizer\n- optimizer for quadratic weighted kappa\n4900+ views, 160+ upvotes

\n\n

[6] Training\n- very simple pytorch training [0.59+]\n8300+ views, 160+ upvotes

\n\n

[7] Inference\n- pytorch inference kernel + lazy TTA\n21000+ views, 360+ upvotes\n- Eye -- Inference(num class=1) ver3\n7000+ views, 140+ upvotes

", "messages": [ { "id": 605288, "postDate": "2019-08-22T08:20:43.377Z", "content": "

[1] All in one\n- Intro APTOS Diabetic Retinopathy (EDA & Starter)\n14000+ views, 350+ upvotes\n- APTOS 2019: DenseNet Keras Starter\n15000+ views, 270+ upvotes\n- APTOS [updated] Albumentation meets Grad-CAM\n2600+ views, 80+ upvotes\n- Basic EDA and baseline pytorch model\n2900+ views, 70+ upvotes\n- Keras+ResNet+Test Time Augmentation\n6000+ views, 50+ upvotes\n- APTOS Blindness Detection - EDA and Keras ResNet50\n2200+ views, 50+ upvotes

\n\n

[2] EDA\n- Image shape distribution - previous and present\n1200+ views, 30+ upvotes\n- APTOS2019: Blindness Detection EDA\n600+ views, 20+ upvotes\n- Basic EDA - train test image distribution check\n600+ views, 10+ upvotes

\n\n

[3] Baseline\n- [APTOS] resnet50 baseline\n10000 views, 140+ upvotes\n- Eye -- EfficientNet Pytorch[LB 0.777] \n5500+ views, 100+ upvotes\n- EfficientNetB5 with Keras (APTOS 2019)\n4300+ views, 90+ upvotes\n- APTOS 2019: Keras Baseline\n5100+ views, 90+ upvotes\n- EfficientNetB4 FastAI - Blindness Detection\n4800+ views, 70+ upvotes

\n\n

[4] Preprocessing\n- APTOS [UpdatedV14] Preprocessing- Ben's & Cropping\n21000+ views, 560+ upvotes\n- Pre-processing train and test images\n4000+ views, 120+ upvotes

\n\n

[5] Optimizer\n- optimizer for quadratic weighted kappa\n4900+ views, 160+ upvotes

\n\n

[6] Training\n- very simple pytorch training [0.59+]\n8300+ views, 160+ upvotes

\n\n

[7] Inference\n- pytorch inference kernel + lazy TTA\n21000+ views, 360+ upvotes\n- Eye -- Inference(num class=1) ver3\n7000+ views, 140+ upvotes

", "rawMarkdown": "[1] All in one\n- [Intro APTOS Diabetic Retinopathy (EDA & Starter)](https://www.kaggle.com/tanlikesmath/intro-aptos-diabetic-retinopathy-eda-starter)\n14000+ views, 350+ upvotes\n- [APTOS 2019: DenseNet Keras Starter](https://www.kaggle.com/xhlulu/aptos-2019-densenet-keras-starter)\n15000+ views, 270+ upvotes\n- [APTOS [updated] Albumentation meets Grad-CAM](https://www.kaggle.com/ratthachat/aptos-updated-albumentation-meets-grad-cam#2.-Prepare-tools-of-the-original-kernel)\n2600+ views, 80+ upvotes\n- [Basic EDA and baseline pytorch model](https://www.kaggle.com/artgor/basic-eda-and-baseline-pytorch-model)\n2900+ views, 70+ upvotes\n- [Keras+ResNet+Test Time Augmentation](https://www.kaggle.com/bharatsingh213/keras-resnet-test-time-augmentation)\n6000+ views, 50+ upvotes\n- [APTOS Blindness Detection - EDA and Keras ResNet50](https://www.kaggle.com/dimitreoliveira/aptos-blindness-detection-eda-and-keras-resnet50)\n2200+ views, 50+ upvotes\n\n[2] EDA\n- [Image shape distribution - previous and present](https://www.kaggle.com/currypurin/image-shape-distribution-previous-and-present)\n1200+ views, 30+ upvotes\n- [APTOS2019: Blindness Detection EDA](https://www.kaggle.com/aleksandradeis/aptos2019-blindness-detection-eda)\n600+ views, 20+ upvotes\n- [Basic EDA - train test image distribution check](https://www.kaggle.com/yangsaewon/basic-eda-train-test-image-distribution-check)\n600+ views, 10+ upvotes\n\n[3] Baseline\n- [[APTOS] resnet50 baseline](https://www.kaggle.com/mathormad/aptos-resnet50-baseline)\n10000 views, 140+ upvotes\n- [Eye -- EfficientNet Pytorch[LB 0.777] ](https://www.kaggle.com/chanhu/eye-efficientnet-pytorch-lb-0-777)\n5500+ views, 100+ upvotes\n- [EfficientNetB5 with Keras (APTOS 2019)](https://www.kaggle.com/carlolepelaars/efficientnetb5-with-keras-aptos-2019)\n4300+ views, 90+ upvotes\n- [APTOS 2019: Keras Baseline](https://www.kaggle.com/ratan123/aptos-2019-keras-baseline)\n5100+ views, 90+ upvotes\n- [EfficientNetB4 FastAI - Blindness Detection](https://www.kaggle.com/hmendonca/efficientnetb4-fastai-blindness-detection)\n4800+ views, 70+ upvotes\n\n[4] Preprocessing\n- [APTOS [UpdatedV14] Preprocessing- Ben's & Cropping](https://www.kaggle.com/ratthachat/aptos-updatedv14-preprocessing-ben-s-cropping)\n21000+ views, 560+ upvotes\n- [Pre-processing train and test images](https://www.kaggle.com/taindow/pre-processing-train-and-test-images)\n4000+ views, 120+ upvotes\n\n[5] Optimizer\n- [optimizer for quadratic weighted kappa](https://www.kaggle.com/abhishek/optimizer-for-quadratic-weighted-kappa)\n4900+ views, 160+ upvotes\n\n[6] Training\n- [very simple pytorch training [0.59+]](https://www.kaggle.com/abhishek/very-simple-pytorch-training-0-59)\n8300+ views, 160+ upvotes\n\n[7] Inference\n- [pytorch inference kernel + lazy TTA](https://www.kaggle.com/abhishek/pytorch-inference-kernel-lazy-tta)\n21000+ views, 360+ upvotes\n- [Eye -- Inference(num class=1) ver3](https://www.kaggle.com/chanhu/eye-inference-num-class-1-ver3)\n7000+ views, 140+ upvotes", "votes": 46 }, { "id": 616688, "postDate": "2019-09-03T11:20:06.067Z", "content": "

Thanks for sharing. I will try to create a new kernel with all these and ensemble all.

", "rawMarkdown": "Thanks for sharing. I will try to create a new kernel with all these and ensemble all.", "votes": 1 }, { "id": 614729, "postDate": "2019-09-01T01:05:15.477Z", "content": "

Awesome list! informative & helpful. Thanks for sharing.

", "rawMarkdown": "Awesome list! informative & helpful. Thanks for sharing.", "votes": 1 }, { "id": 608522, "postDate": "2019-08-26T21:42:23.523Z", "content": "

Thanks for sharing my work, for me, this competition was a great opportunity to learn, lot's of informative content all around.

", "rawMarkdown": "Thanks for sharing my work, for me, this competition was a great opportunity to learn, lot's of informative content all around.", "votes": 1 }, { "id": 605457, "postDate": "2019-08-22T11:41:36.373Z", "content": "

Great summarization of all kinds of kernels, thanks!

", "rawMarkdown": "Great summarization of all kinds of kernels, thanks!", "votes": 1 }, { "id": 609908, "postDate": "2019-08-28T08:10:17.440Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 609900, "postDate": "2019-08-28T07:55:54.297Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 609716, "postDate": "2019-08-28T03:12:42.610Z", "content": "

Thanks!! Is there kernel to use data augmentation to deal with the data imbalance? I want to preprocess locally for balancing and preprocessing.

", "rawMarkdown": "Thanks!! Is there kernel to use data augmentation to deal with the data imbalance? I want to preprocess locally for balancing and preprocessing. ", "votes": 1, "isDeleted": true }, { "id": 620057, "postDate": "2019-09-07T00:26:56.903Z", "content": "

Thank you for your efforts

", "rawMarkdown": "Thank you for your efforts", "votes": 1 }, { "id": 618580, "postDate": "2019-09-05T10:19:59.227Z", "content": "

nice resource. thank you

", "rawMarkdown": "nice resource. thank you", "votes": 1 }, { "id": 609449, "postDate": "2019-08-27T18:18:44.527Z", "content": "

this looks good, thanks !

", "rawMarkdown": "this looks good, thanks !", "votes": 1 }, { "id": 607997, "postDate": "2019-08-26T07:30:33.167Z", "content": "

Thanks, this helped a lot

", "rawMarkdown": "Thanks, this helped a lot\n", "votes": 1 }, { "id": 605294, "postDate": "2019-08-22T08:24:54.960Z", "content": "

wow very informative ! many thanks

", "rawMarkdown": "wow very informative ! many thanks", "votes": 1 } ], "comments": [ { "id": 616688, "author_name": "Hashim", "author_url": "", "post_date": "2019-09-03T11:20:06.067000", "content": "

Thanks for sharing. I will try to create a new kernel with all these and ensemble all.

", "votes": 1, "replies": [] }, { "id": 614729, "author_name": "Youngsoo Lee", "author_url": "", "post_date": "2019-09-01T01:05:15.477000", "content": "

Awesome list! informative & helpful. Thanks for sharing.

", "votes": 1, "replies": [] }, { "id": 608522, "author_name": "DimitreOliveira", "author_url": "", "post_date": "2019-08-26T21:42:23.523000", "content": "

Thanks for sharing my work, for me, this competition was a great opportunity to learn, lot's of informative content all around.

", "votes": 1, "replies": [] }, { "id": 605457, "author_name": "Liu.Siyuan", "author_url": "", "post_date": "2019-08-22T11:41:36.373000", "content": "

Great summarization of all kinds of kernels, thanks!

", "votes": 1, "replies": [] }, { "id": 609908, "author_name": "", "author_url": "", "post_date": "2019-08-28T08:10:17.440000", "content": "", "votes": 1, "replies": [] }, { "id": 609900, "author_name": "", "author_url": "", "post_date": "2019-08-28T07:55:54.297000", "content": "", "votes": 1, "replies": [] }, { "id": 609716, "author_name": "", "author_url": "", "post_date": "2019-08-28T03:12:42.610000", "content": "

Thanks!! Is there kernel to use data augmentation to deal with the data imbalance? I want to preprocess locally for balancing and preprocessing.

", "votes": 1, "replies": [] }, { "id": 620057, "author_name": "kodakku", "author_url": "", "post_date": "2019-09-07T00:26:56.903000", "content": "

Thank you for your efforts

", "votes": 1, "replies": [] }, { "id": 618580, "author_name": "ls", "author_url": "", "post_date": "2019-09-05T10:19:59.227000", "content": "

nice resource. thank you

", "votes": 1, "replies": [] }, { "id": 609449, "author_name": "Anurag Trivedi", "author_url": "", "post_date": "2019-08-27T18:18:44.527000", "content": "

this looks good, thanks !

", "votes": 1, "replies": [] }, { "id": 607997, "author_name": "Harsh Sharma", "author_url": "", "post_date": "2019-08-26T07:30:33.167000", "content": "

Thanks, this helped a lot

", "votes": 1, "replies": [] }, { "id": 605294, "author_name": "leixiang@AInnovation", "author_url": "", "post_date": "2019-08-22T08:24:54.960000", "content": "

wow very informative ! many thanks

", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "605288": "[1] All in one\n- [Intro APTOS Diabetic Retinopathy (EDA & Starter)](https://www.kaggle.com/tanlikesmath/intro-aptos-diabetic-retinopathy-eda-starter)\n14000+ views, 350+ upvotes\n- [APTOS 2019: DenseNet Keras Starter](https://www.kaggle.com/xhlulu/aptos-2019-densenet-keras-starter)\n15000+ views, 270+ upvotes\n- [APTOS [updated] Albumentation meets Grad-CAM](https://www.kaggle.com/ratthachat/aptos-updated-albumentation-meets-grad-cam#2.-Prepare-tools-of-the-original-kernel)\n2600+ views, 80+ upvotes\n- [Basic EDA and baseline pytorch model](https://www.kaggle.com/artgor/basic-eda-and-baseline-pytorch-model)\n2900+ views, 70+ upvotes\n- [Keras+ResNet+Test Time Augmentation](https://www.kaggle.com/bharatsingh213/keras-resnet-test-time-augmentation)\n6000+ views, 50+ upvotes\n- [APTOS Blindness Detection - EDA and Keras ResNet50](https://www.kaggle.com/dimitreoliveira/aptos-blindness-detection-eda-and-keras-resnet50)\n2200+ views, 50+ upvotes\n\n[2] EDA\n- [Image shape distribution - previous and present](https://www.kaggle.com/currypurin/image-shape-distribution-previous-and-present)\n1200+ views, 30+ upvotes\n- [APTOS2019: Blindness Detection EDA](https://www.kaggle.com/aleksandradeis/aptos2019-blindness-detection-eda)\n600+ views, 20+ upvotes\n- [Basic EDA - train test image distribution check](https://www.kaggle.com/yangsaewon/basic-eda-train-test-image-distribution-check)\n600+ views, 10+ upvotes\n\n[3] Baseline\n- [[APTOS] resnet50 baseline](https://www.kaggle.com/mathormad/aptos-resnet50-baseline)\n10000 views, 140+ upvotes\n- [Eye -- EfficientNet Pytorch[LB 0.777] ](https://www.kaggle.com/chanhu/eye-efficientnet-pytorch-lb-0-777)\n5500+ views, 100+ upvotes\n- [EfficientNetB5 with Keras (APTOS 2019)](https://www.kaggle.com/carlolepelaars/efficientnetb5-with-keras-aptos-2019)\n4300+ views, 90+ upvotes\n- [APTOS 2019: Keras Baseline](https://www.kaggle.com/ratan123/aptos-2019-keras-baseline)\n5100+ views, 90+ upvotes\n- [EfficientNetB4 FastAI - Blindness Detection](https://www.kaggle.com/hmendonca/efficientnetb4-fastai-blindness-detection)\n4800+ views, 70+ upvotes\n\n[4] Preprocessing\n- [APTOS [UpdatedV14] Preprocessing- Ben's & Cropping](https://www.kaggle.com/ratthachat/aptos-updatedv14-preprocessing-ben-s-cropping)\n21000+ views, 560+ upvotes\n- [Pre-processing train and test images](https://www.kaggle.com/taindow/pre-processing-train-and-test-images)\n4000+ views, 120+ upvotes\n\n[5] Optimizer\n- [optimizer for quadratic weighted kappa](https://www.kaggle.com/abhishek/optimizer-for-quadratic-weighted-kappa)\n4900+ views, 160+ upvotes\n\n[6] Training\n- [very simple pytorch training [0.59+]](https://www.kaggle.com/abhishek/very-simple-pytorch-training-0-59)\n8300+ views, 160+ upvotes\n\n[7] Inference\n- [pytorch inference kernel + lazy TTA](https://www.kaggle.com/abhishek/pytorch-inference-kernel-lazy-tta)\n21000+ views, 360+ upvotes\n- [Eye -- Inference(num class=1) ver3](https://www.kaggle.com/chanhu/eye-inference-num-class-1-ver3)\n7000+ views, 140+ upvotes", "616688": "Thanks for sharing. I will try to create a new kernel with all these and ensemble all.", "614729": "Awesome list! informative & helpful. Thanks for sharing.", "608522": "Thanks for sharing my work, for me, this competition was a great opportunity to learn, lot's of informative content all around.", "605457": "Great summarization of all kinds of kernels, thanks!", "609908": "", "609900": "", "609716": "Thanks!! Is there kernel to use data augmentation to deal with the data imbalance? I want to preprocess locally for balancing and preprocessing. ", "620057": "Thank you for your efforts", "618580": "nice resource. thank you", "609449": "this looks good, thanks !", "607997": "Thanks, this helped a lot\n", "605294": "wow very informative ! many thanks" } }