![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1554318%2Fffd3fc8d02f285bed1446e60939a793b%2F2020-03-16%209.27.22.png?generation=1584408319262779&alt=media\")
Post links to your external data sources here before the deadline specified in the rules.
", "messages": [ { "id": 698947, "postDate": "2019-12-19T23:54:32.743Z", "content": "Post links to your external data sources here before the deadline specified in the rules.
", "rawMarkdown": "Post links to your external data sources here before the deadline specified in the rules.", "votes": 15 }, { "id": 709666, "postDate": "2020-01-03T19:45:27.173Z", "content": "cadene pretarined models\nhttps://github.com/Cadene/pretrained-models.pytorch
torchvision.models\nhttps://pytorch.org/docs/stable/torchvision/models.html
Pytorch EfficientNet\nhttps://github.com/lukemelas/EfficientNet-PyTorch
pytorch image models\nhttps://github.com/rwightman/pytorch-image-models/tree/master/timm
\n\npytorch pretrained models\nhttps://github.com/Cadene/pretrained-models.pytorch
\n\nekush dataset\nhttps://shahariarrabby.github.io/ekush/#overview
\n\ntorchvision.models\nhttps://pytorch.org/docs/stable/torchvision/models.html
\n\nPytorch EfficientNet\nhttps://github.com/lukemelas/EfficientNet-PyTorch
", "rawMarkdown": "pytorch image models\nhttps://github.com/rwightman/pytorch-image-models/tree/master/timm\n\npytorch pretrained models\nhttps://github.com/Cadene/pretrained-models.pytorch\n\nekush dataset\nhttps://shahariarrabby.github.io/ekush/#overview\n\ntorchvision.models\nhttps://pytorch.org/docs/stable/torchvision/models.html\n\nPytorch EfficientNet\nhttps://github.com/lukemelas/EfficientNet-PyTorch\n", "votes": 1 }, { "id": 743631, "postDate": "2020-02-12T06:39:18.590Z", "content": "I plan to experiemen with the following as posted by @ipythonx (since I dont see explicit links mentioned here just in case):
\n\nBanglaLekha-Isolated-Data-Set\n Ekush-Data-Set \nNumtaDB\nISI Image Data \nCMATERdb
", "rawMarkdown": "I plan to experiemen with the following as posted by @ipythonx (since I dont see explicit links mentioned here just in case): \n\n[ BanglaLekha-Isolated-Data-Set](https://data.mendeley.com/datasets/hf6sf8zrkc/2?__cf_chl_captcha_tk__=0eef3f155bb8cdbd4318462de9ce43b06e8b36fe-1576926645-0-ARwIJOlc0pqWADHpP7QJNSnlLmWPdU95TWcw8RABhCTk-MHuKadBfXUxQednw2omNlWNN2YFffaQjm6HFtzYhbKIaX6ZXIydnz6nMJUQL1p3MoCjLaD8C1Y7P2F33-MAWE3Eo1NstIjhrEzpMBoz4nmc0z8Fx1uftg5rDt5YCNCiWsBY4VcaijT_NXaReFueYu3CBgk-he-9c_Xr7FhsPoQCpg8ThfHCmELrGY8f5EKmi5eFhBEWP6-1JpByjl4Z4Qr5-3zWmjn0uBsoPF2DKD6SE10QnF1hcUP5rHbSfV4Iv8mLUYoTLLeArXclMmeNEQuWKo_AgS5fAzLFuFCf_DI)\n [Ekush-Data-Set](https://shahariarrabby.github.io/ekush/#overview) \n[NumtaDB](https://www.kaggle.com/BengaliAI/numta)\n[ISI Image Data](https://www.isical.ac.in/~ujjwal/download/BanglaNumeral.html) \n[CMATERdb](https://code.google.com/archive/p/cmaterdb/)\n", "votes": 1, "replies": [ { "id": 743652, "postDate": "2020-02-12T06:53:02.603Z", "content": "NumtaDB is only about bengali digit (0,1,2,..9), so you should skip this). ISI is free but need official invitation to the author to get, so bit trouble. I think if you really plan so, you should go with Ekush and BanglaLekha, more than enough.
", "rawMarkdown": "NumtaDB is only about bengali digit (0,1,2,..9), so you should skip this). ISI is free but need official invitation to the author to get, so bit trouble. I think if you really plan so, you should go with Ekush and BanglaLekha, more than enough. ", "votes": 6 }, { "id": 744525, "postDate": "2020-02-12T22:47:47.610Z", "content": "I see, thank you for the information!
", "rawMarkdown": "I see, thank you for the information!", "votes": 1 } ] }, { "id": 737174, "postDate": "2020-02-05T02:35:29.800Z", "content": "If I create my own labelled dataset and train the model on it, then do I have to disclose it / how do I disclose it?\nWhat if I don't disclose it - how can the organisers possibly come to know, esp. if the model is trained offline?\n@sohier
", "rawMarkdown": "If I create my own labelled dataset and train the model on it, then do I have to disclose it / how do I disclose it?\nWhat if I don't disclose it - how can the organisers possibly come to know, esp. if the model is trained offline?\n@sohier ", "votes": 1, "replies": [ { "id": 738404, "postDate": "2020-02-06T13:33:39.410Z", "content": "i think they do hard checking only for those who finish on high,imagine you end up being in gold zone or in prize zone,then kaggle will invest more time on investigating you and your model,if they can catch you then possibly your account can be removed permanently like this : https://www.kaggle.com/c/petfinder-adoption-prediction/discussion/125436#715470 ,they take very good care for the people who finish on high,if you stay far away from gold zone in private lb then maybe they won't catch it but if they can detect that you are cheating then you are in immense trouble!
", "rawMarkdown": "i think they do hard checking only for those who finish on high,imagine you end up being in gold zone or in prize zone,then kaggle will invest more time on investigating you and your model,if they can catch you then possibly your account can be removed permanently like this : https://www.kaggle.com/c/petfinder-adoption-prediction/discussion/125436#715470 ,they take very good care for the people who finish on high,if you stay far away from gold zone in private lb then maybe they won't catch it but if they can detect that you are cheating then you are in immense trouble!", "votes": 2 } ] }, { "id": 757894, "postDate": "2020-02-27T08:09:43.723Z", "content": "cadene pretarined models\nhttps://github.com/Cadene/pretrained-models.pytorch
\n\nPytorch EfficientNet\nhttps://github.com/lukemelas/EfficientNet-PyTorch
", "rawMarkdown": "cadene pretarined models\nhttps://github.com/Cadene/pretrained-models.pytorch\n\nPytorch EfficientNet\nhttps://github.com/lukemelas/EfficientNet-PyTorch", "votes": 2 }, { "id": 767565, "postDate": "2020-03-09T20:20:36.080Z", "content": "https://github.com/MinhasKamal/BengaliDictionary
", "rawMarkdown": "https://github.com/MinhasKamal/BengaliDictionary" }, { "id": 700602, "postDate": "2019-12-22T09:27:41.947Z", "content": "EfficientNet Keras\nhttps://github.com/qubvel/efficientnet
", "rawMarkdown": "EfficientNet Keras\nhttps://github.com/qubvel/efficientnet", "votes": 2 }, { "id": 775379, "postDate": "2020-03-16T16:13:28.933Z", "content": "@vibhore94 Oh sir this is magic...hope I found it sooner :)
", "rawMarkdown": "@vibhore94 Oh sir this is magic...hope I found it sooner :)", "votes": 1 }, { "id": 775575, "postDate": "2020-03-16T20:25:30.547Z", "content": "too late though. deadline for external data was 7 days ago
", "rawMarkdown": "too late though. deadline for external data was 7 days ago", "votes": 6 }, { "id": 775593, "postDate": "2020-03-16T20:57:46.713Z", "content": "Wow, and some of them were not declared by someone before...\nBut I saw some cases that winner's solution includes extra data without disclosing it before the deadline and it seems kaggle and host doesn't check it seriously...\nBut I think they should do it more vigorously.
", "rawMarkdown": "Wow, and some of them were not declared by someone before...\nBut I saw some cases that winner's solution includes extra data without disclosing it before the deadline and it seems kaggle and host doesn't check it seriously...\nBut I think they should do it more vigorously.", "votes": 2 }, { "id": 775594, "postDate": "2020-03-16T20:59:40.430Z", "content": "But if he would have posted that he uses GANs, all his advantage would have been gone? Kalpurush was posted before, but not all yeah. Hard to judge.
", "rawMarkdown": "But if he would have posted that he uses GANs, all his advantage would have been gone? Kalpurush was posted before, but not all yeah. Hard to judge.", "votes": 3 }, { "id": 775628, "postDate": "2020-03-16T22:21:04.873Z", "content": "Yes, as same as other rules it's not crystal clear. The question about which kind of data is considered as external data and which is not should be judged by host.
\n\nI'm not sure at all that the GANs are considered as external data or not, but if it's external data and if it helps to improve score, I guess that's why this rule exists. That advantage should have gone a week ago.\n(Again, I'm not judging he is violating the rule or not. It should be done by host. And also he may not use these in his final solutions.)
", "rawMarkdown": "Yes, as same as other rules it's not crystal clear. The question about which kind of data is considered as external data and which is not should be judged by host. \n\nI'm not sure at all that the GANs are considered as external data or not, but if it's external data and if it helps to improve score, I guess that's why this rule exists. That advantage should have gone a week ago.\n(Again, I'm not judging he is violating the rule or not. It should be done by host. And also he may not use these in his final solutions.)", "votes": 2 }, { "id": 775800, "postDate": "2020-03-17T01:10:29.067Z", "content": "I'm sorry. I am new to Kaggle and did not understand the rules for publishing external data.\nCan I ask where the exact deadline could be found?
", "rawMarkdown": "\nI'm sorry. I am new to Kaggle and did not understand the rules for publishing external data.\nCan I ask where the exact deadline could be found?", "votes": 1 }, { "id": 775838, "postDate": "2020-03-17T01:35:02.273Z", "content": "Here it is.(in C.)\nI totally understand what you're saying, it should be posted on top of this thread. Most of competitors know just because it's always same as team merge deadline.\nIn my honest opinion it's pretty minor thing compared to what you achieved. Congrats and you definitely deserve it!\n
https://www.kaggle.com/c/bengaliai-cv19/rules
", "rawMarkdown": "Here it is.(in C.)\nI totally understand what you're saying, it should be posted on top of this thread. Most of competitors know just because it's always same as team merge deadline.\nIn my honest opinion it's pretty minor thing compared to what you achieved. Congrats and you definitely deserve it!\n\n\nhttps://www.kaggle.com/c/bengaliai-cv19/rules", "votes": 6 } ] }, { "id": 767580, "postDate": "2020-03-09T20:41:12.393Z", "content": "I have created a set of synthetic characters using the following Bengali fonts:\nEkushey_Durga-Normal.ttf\nEkushey_Sharifa-Normal.ttf\nEkushey_Puja-Normal.ttf\nGalada-Regular.ttf\nEkushey_Punarbhaba-Normal.ttf\nBenSenHandwriting.ttf\nEkushey_Godhuli-Normal.ttf\nEkushey_Sumit-Normal.ttf\nSiyamrupali.ttf\nAtma-Bold.ttf\nBaloo_Da-Regular.ttf\nEkushey_Mohua-Normal.ttf\nKalpurush-Regular.ttf
", "rawMarkdown": "I have created a set of synthetic characters using the following Bengali fonts:\nEkushey_Durga-Normal.ttf\nEkushey_Sharifa-Normal.ttf\nEkushey_Puja-Normal.ttf\nGalada-Regular.ttf\nEkushey_Punarbhaba-Normal.ttf\nBenSenHandwriting.ttf\nEkushey_Godhuli-Normal.ttf\nEkushey_Sumit-Normal.ttf\nSiyamrupali.ttf\nAtma-Bold.ttf\nBaloo_Da-Regular.ttf\nEkushey_Mohua-Normal.ttf\nKalpurush-Regular.ttf", "replies": [ { "id": 772346, "postDate": "2020-03-15T11:40:26.513Z", "content": "Also created a set of synthetic characters using the following fonts:\nAmarBangl, Bensenhandwriting, CHITMI, ChitraMJ, Durga, JugantorMJ, kalpurush, KeertankhuIaM, KumarkhaliMJ, Kumarkhali, Mohua, Nikosh, PANDM, Puja, PunarBhaba, Sharifa, Siyamrupali, SolaimanLi, SUTOM, SutonnyOM, TEESM
", "rawMarkdown": "Also created a set of synthetic characters using the following fonts:\nAmarBangl, Bensenhandwriting, CHITMI, ChitraMJ, Durga, JugantorMJ, kalpurush, KeertankhuIaM, KumarkhaliMJ, Kumarkhali, Mohua, Nikosh, PANDM, Puja, PunarBhaba, Sharifa, Siyamrupali, SolaimanLi, SUTOM, SutonnyOM, TEESM" }, { "id": 775534, "postDate": "2020-03-16T19:20:11.463Z", "content": "I will just add more fonts:
\n\nhttp://www.freebanglafont.com/download.php?id=649\nhttp://www.freebanglafont.com/download.php?id=762\nhttps://www.freeforfonts.com/kohinoor-bangla-font-family/\nhttp://www.freebanglafont.com/download.php?id=765
\n\nhttp://www.freebanglafont.com/download.php?id=659\nhttps://www.stylemyname.com/Chit-Mi\nhttp://www.freebanglafont.com/download.php?id=708\nhttp://www.freebanglafont.com/download.php?id=724\nhttp://www.freebanglafont.com/download.php?id=769\nhttps://blogfonts.com/ekushey-punarbhaba.font\nhttp://www.freebanglafont.com/download.php?id=658\nhttp://www.freebanglafont.com/download.php?id=700\nhttp://www.freebanglafont.com/download.php?id=690\nhttp://www.freebanglafont.com/download.php?id=763\nhttp://www.freebanglafont.com/download.php?id=708\n.TTF\">https://github.com/torifat/bangla-font/blob/master/src/fonts/SUTOM_.TTF\nhttp://www.freebanglafont.com/download.php?id=651
", "rawMarkdown": "I will just add more fonts:\n\nhttp://www.freebanglafont.com/download.php?id=649\nhttp://www.freebanglafont.com/download.php?id=762\nhttps://www.freeforfonts.com/kohinoor-bangla-font-family/\nhttp://www.freebanglafont.com/download.php?id=765\n\nhttp://www.freebanglafont.com/download.php?id=659\nhttps://www.stylemyname.com/Chit-Mi\nhttp://www.freebanglafont.com/download.php?id=708\nhttp://www.freebanglafont.com/download.php?id=724\nhttp://www.freebanglafont.com/download.php?id=769\nhttps://blogfonts.com/ekushey-punarbhaba.font\nhttp://www.freebanglafont.com/download.php?id=658\nhttp://www.freebanglafont.com/download.php?id=700\nhttp://www.freebanglafont.com/download.php?id=690\nhttp://www.freebanglafont.com/download.php?id=763\nhttp://www.freebanglafont.com/download.php?id=708\nhttps://github.com/torifat/bangla-font/blob/master/src/fonts/SUTOM___.TTF\nhttp://www.freebanglafont.com/download.php?id=651\n" }, { "id": 775542, "postDate": "2020-03-16T19:35:37.677Z", "content": "Looks like you guys Learned the Magic! Grats!
", "rawMarkdown": "Looks like you guys Learned the Magic! Grats!", "votes": 1 }, { "id": 775572, "postDate": "2020-03-16T20:21:53.463Z", "rawMarkdown": "", "isDeleted": true }, { "id": 775573, "postDate": "2020-03-16T20:23:39.123Z", "content": "… but you are not allowed to use them in your solution. The deadline was 1 week ago: March 9, 2020.
", "rawMarkdown": "… but you are not allowed to use them in your solution. The deadline was 1 week ago: March 9, 2020.", "votes": 5 } ] }, { "id": 775519, "postDate": "2020-03-16T19:03:20.167Z", "content": "https://www.kaggle.com/pkugoodspeed/bengali-competition-checkpoints\nThanks
", "rawMarkdown": "https://www.kaggle.com/pkugoodspeed/bengali-competition-checkpoints\nThanks" }, { "id": 775464, "postDate": "2020-03-16T17:48:42.803Z", "content": "https://keras.io/applications/#densenet
", "rawMarkdown": "https://keras.io/applications/#densenet" }, { "id": 775141, "postDate": "2020-03-16T10:59:59.630Z", "content": "https://github.com/rwightman/gen-efficientnet-pytorch
", "rawMarkdown": "https://github.com/rwightman/gen-efficientnet-pytorch" }, { "id": 774937, "postDate": "2020-03-16T04:43:38.647Z", "content": "https://github.com/Callidior/keras-applications/releases/download/efficientnet\nhttps://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b4_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5\n https://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b3_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5\n https://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b5_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5
", "rawMarkdown": "https://github.com/Callidior/keras-applications/releases/download/efficientnet\nhttps://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b4_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5\n https://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b3_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5\n https://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b5_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5" }, { "id": 771819, "postDate": "2020-03-14T16:59:33.120Z", "content": "https://www.kaggle.com/yeayates21/bengaliaitrainedmodelv4
", "rawMarkdown": "https://www.kaggle.com/yeayates21/bengaliaitrainedmodelv4" }, { "id": 767524, "postDate": "2020-03-09T18:41:58.873Z", "content": "https://www.kaggle.com/lucky227/ooooo3\nhttps://www.kaggle.com/rsmits/kerasefficientnetb3\nhttps://www.kaggle.com/h030162/se-resnext50-32x4d-fold2\nhttps://www.kaggle.com/h030162/se-resnext50-32x4d-fold3
", "rawMarkdown": "https://www.kaggle.com/lucky227/ooooo3\nhttps://www.kaggle.com/rsmits/kerasefficientnetb3\nhttps://www.kaggle.com/h030162/se-resnext50-32x4d-fold2\nhttps://www.kaggle.com/h030162/se-resnext50-32x4d-fold3" }, { "id": 767427, "postDate": "2020-03-09T16:07:00.643Z", "content": "https://github.com/adobe/antialiased-cnns
", "rawMarkdown": "https://github.com/adobe/antialiased-cnns" }, { "id": 767309, "postDate": "2020-03-09T13:18:17.530Z", "content": "http://places2.csail.mit.edu/index.html
", "rawMarkdown": "http://places2.csail.mit.edu/index.html" }, { "id": 767290, "postDate": "2020-03-09T12:49:11.197Z", "content": "cadene pre-trained models: https://github.com/Cadene/pretrained-models.pytorch\ntorchvision.models: https://pytorch.org/docs/stable/torchvision/models.html
", "rawMarkdown": "cadene pre-trained models: https://github.com/Cadene/pretrained-models.pytorch\ntorchvision.models: https://pytorch.org/docs/stable/torchvision/models.html" }, { "id": 767041, "postDate": "2020-03-09T04:32:23.433Z", "content": "I may use: https://github.com/microsoft/computervision-recipes
", "rawMarkdown": "I may use: https://github.com/microsoft/computervision-recipes" }, { "id": 766892, "postDate": "2020-03-08T22:53:47.007Z", "content": "https://github.com/Cadene/pretrained-models.pytorch\nhttps://github.com/lukemelas/EfficientNet-PyTorch\nhttps://shahariarrabby.github.io/ekush/\nhttp://www.openslr.org/37/\nhttp://www.openslr.org/53/
", "rawMarkdown": "https://github.com/Cadene/pretrained-models.pytorch\nhttps://github.com/lukemelas/EfficientNet-PyTorch\nhttps://shahariarrabby.github.io/ekush/\nhttp://www.openslr.org/37/\nhttp://www.openslr.org/53/" }, { "id": 766389, "postDate": "2020-03-08T04:40:03.673Z", "content": "maybe I will try to somehow use this (it's chinese):
\n\nhttp://www.nlpr.ia.ac.cn/databases/handwriting/Home.html
\n\nhttp://www.nlpr.ia.ac.cn/databases/handwriting/Download.html
\n\nhttp://www.nlpr.ia.ac.cn/databases/Download/feature_data/1.0train-gb1.rar
\n\nhttp://www.nlpr.ia.ac.cn/databases/Download/feature_data/1.0test-gb1.rar
", "rawMarkdown": "maybe I will try to somehow use this (it's chinese):\n\nhttp://www.nlpr.ia.ac.cn/databases/handwriting/Home.html\n\nhttp://www.nlpr.ia.ac.cn/databases/handwriting/Download.html\n\nhttp://www.nlpr.ia.ac.cn/databases/Download/feature_data/1.0train-gb1.rar\n\nhttp://www.nlpr.ia.ac.cn/databases/Download/feature_data/1.0test-gb1.rar" }, { "id": 766145, "postDate": "2020-03-07T18:22:57.073Z", "content": "https://github.com/OverLordGoldDragon/keras-adamw - AdamW for Keras\nhttps://github.com/albu/albumentations - Albumentations library for data augmentation
", "rawMarkdown": "https://github.com/OverLordGoldDragon/keras-adamw - AdamW for Keras\nhttps://github.com/albu/albumentations - Albumentations library for data augmentation" }, { "id": 765860, "postDate": "2020-03-07T08:30:05.017Z", "content": "PyTorch Image Models\nhttps://github.com/rwightman/pytorch-image-models
", "rawMarkdown": "PyTorch Image Models\nhttps://github.com/rwightman/pytorch-image-models" }, { "id": 765584, "postDate": "2020-03-06T20:34:24.887Z", "content": "cadene pretarined models\nhttps://github.com/Cadene/pretrained-models.pytorch
", "rawMarkdown": "cadene pretarined models\nhttps://github.com/Cadene/pretrained-models.pytorch" }, { "id": 761313, "postDate": "2020-03-02T11:34:51.427Z", "content": "https://pypi.org/project/efficientnet/
", "rawMarkdown": "https://pypi.org/project/efficientnet/" }, { "id": 760527, "postDate": "2020-03-01T12:17:10.867Z", "content": "https://github.com/Cadene/pretrained-models.pytorch\nhttps://github.com/lukemelas/EfficientNet-PyTorch\nhttps://github.com/osmr/imgclsmob\nhttps://github.com/qubvel/segmentation_models.pytorch
", "rawMarkdown": "https://github.com/Cadene/pretrained-models.pytorch\nhttps://github.com/lukemelas/EfficientNet-PyTorch\nhttps://github.com/osmr/imgclsmob\nhttps://github.com/qubvel/segmentation_models.pytorch" }, { "id": 760345, "postDate": "2020-03-01T06:37:31.903Z", "content": "EfficientNet models from Tensorflow repo https://github.com/tensorflow/tpu/tree/69a06b6a11ff9ba8835b27348f66f33864af93fe/models/official/efficientnet https://github.com/Kyubyong/wordvectors\nhttps://pypi.org/project/grapheme/
", "rawMarkdown": "EfficientNet models from Tensorflow repo https://github.com/tensorflow/tpu/tree/69a06b6a11ff9ba8835b27348f66f33864af93fe/models/official/efficientnet https://github.com/Kyubyong/wordvectors\nhttps://pypi.org/project/grapheme/" }, { "id": 759979, "postDate": "2020-02-29T17:24:59.090Z", "content": "https://pypi.org/project/efficientnet/
", "rawMarkdown": "https://pypi.org/project/efficientnet/" }, { "id": 759879, "postDate": "2020-02-29T15:17:08.263Z", "content": "https://github.com/aitorzip/PyTorch-CycleGAN
", "rawMarkdown": "https://github.com/aitorzip/PyTorch-CycleGAN" }, { "id": 759852, "postDate": "2020-02-29T14:47:33.987Z", "content": "Pretrained models: https://github.com/osmr/imgclsmob
", "rawMarkdown": "Pretrained models: https://github.com/osmr/imgclsmob" }, { "id": 759847, "postDate": "2020-02-29T14:42:07.583Z", "content": "keras-efficientnets\nhttps://pypi.org/project/keras-efficientnets/
", "rawMarkdown": "keras-efficientnets\nhttps://pypi.org/project/keras-efficientnets/" }, { "id": 759304, "postDate": "2020-02-28T21:40:21.327Z", "content": "https://github.com/tensorflow/models/\nhttps://github.com/qubvel/classification_models/tree/master/classification_models/models\nhttps://www.kaggle.com/rsmits/kerasefficientnetb3\nhttps://github.com/qubvel/efficientnet
", "rawMarkdown": "https://github.com/tensorflow/models/\nhttps://github.com/qubvel/classification_models/tree/master/classification_models/models\nhttps://www.kaggle.com/rsmits/kerasefficientnetb3\nhttps://github.com/qubvel/efficientnet" }, { "id": 755231, "postDate": "2020-02-24T15:25:04.997Z", "content": "Should we disclosure private scripts and libraries we use as well? I have a couple of packages I use as helpers. (Some of them are available on public GitHub, while others are not). Should I list them here as well? None of them includes any pre-trained weights or training data.
", "rawMarkdown": "Should we disclosure private scripts and libraries we use as well? I have a couple of packages I use as helpers. (Some of them are available on public GitHub, while others are not). Should I list them here as well? None of them includes any pre-trained weights or training data." }, { "id": 749097, "postDate": "2020-02-18T09:42:15.683Z", "content": "Is it possible to use generated dataset? Symbols generated by draw it on context (like here)
", "rawMarkdown": "Is it possible to use generated dataset? Symbols generated by draw it on context (like [here](https://www.kaggle.com/kaushal2896/bengali-graphemes-starter-eda-multi-output-cnn))" }, { "id": 748570, "postDate": "2020-02-17T17:13:28.663Z", "content": "I don't see any explicit mention of the transfer learning models available via keras/tensorflow, so I figured I'd post them here just in case they'd be of help to anyone (and so I can use them 😃).
\n\nhttps://www.tensorflow.org/api_docs/python/tf/keras/applications
", "rawMarkdown": "I don't see any explicit mention of the transfer learning models available via keras/tensorflow, so I figured I'd post them here just in case they'd be of help to anyone (and so I can use them 😃).\n\nhttps://www.tensorflow.org/api_docs/python/tf/keras/applications" }, { "id": 746389, "postDate": "2020-02-14T23:43:36.660Z", "content": "If someone has already has posted an external data set that I am using, am I required to also post it as a disclosure?
\n\n(I am using torchvision.models found here https://pytorch.org/docs/stable/torchvision/models.html)
No, duplicative disclosures aren't necessary.
", "rawMarkdown": "No, duplicative disclosures aren't necessary." } ] }, { "id": 740745, "postDate": "2020-02-09T18:14:27.070Z", "content": "Are we allowed to install new packages in the kaggle env ?
\n\nLike in : https://www.kaggle.com/c/severstal-steel-defect-detection/discussion/109679 ?
\n\nIt will be a lot easier than to copy past code into a single script.
", "rawMarkdown": "Are we allowed to install new packages in the kaggle env ? \n\nLike in : https://www.kaggle.com/c/severstal-steel-defect-detection/discussion/109679 ? \n\nIt will be a lot easier than to copy past code into a single script.\n" }, { "id": 740345, "postDate": "2020-02-09T09:49:13.297Z", "content": "keras pretrained from https://github.com/qubvel/classification_models
", "rawMarkdown": "keras pretrained from https://github.com/qubvel/classification_models" }, { "id": 740059, "postDate": "2020-02-08T19:25:16.100Z", "content": "Using pretraining on https://www.kaggle.com/rishianand/devanagari-character-set
", "rawMarkdown": "Using pretraining on https://www.kaggle.com/rishianand/devanagari-character-set" }, { "id": 734750, "postDate": "2020-02-02T00:31:51.103Z", "content": "Is that fair that I train the model using a huge amount of resources, then save and upload the weights (.h5) to the kernel and make inference only?
", "rawMarkdown": "Is that fair that I train the model using a huge amount of resources, then save and upload the weights (.h5) to the kernel and make inference only?", "replies": [ { "id": 735796, "postDate": "2020-02-03T13:20:49.177Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 725713, "postDate": "2020-01-22T12:27:11.790Z", "content": "if I train a generative model offline for data augmentation and use the augmented data in a kaggle kernel, what do i need to publish here?
", "rawMarkdown": "if I train a generative model offline for data augmentation and use the augmented data in a kaggle kernel, what do i need to publish here? \n", "replies": [ { "id": 725981, "postDate": "2020-01-22T17:27:14.430Z", "content": "As long as the generative model is trained on the competition data you wouldn't need to disclose anything else. Your augmented data is your work, like your code.
", "rawMarkdown": "As long as the generative model is trained on the competition data you wouldn't need to disclose anything else. Your augmented data is your work, like your code. " } ] }, { "id": 722356, "postDate": "2020-01-18T13:06:07.193Z", "content": "As I assumed this competition allow me to use pretrained models and just make predictions using test data via kernel here at Kaggle for submitting. Can you tell my what is forbidden? Should I make pretrained weights public or something something else? I'm newbie in this thread, don't get me wrong please.
", "rawMarkdown": "As I assumed this competition allow me to use pretrained models and just make predictions using test data via kernel here at Kaggle for submitting. Can you tell my what is forbidden? Should I make pretrained weights public or something something else? I'm newbie in this thread, don't get me wrong please.", "replies": [ { "id": 725982, "postDate": "2020-01-22T17:27:47.057Z", "content": "A link to the source of the pretrained weights is fine.
", "rawMarkdown": "A link to the source of the pretrained weights is fine." }, { "id": 727578, "postDate": "2020-01-23T20:53:43.533Z", "content": "Understood.\nWhat about external packages like tensorflow-addons?\nI would like to use it but this package is not installed yet, also competition conditions don't allow me to use internet for installing anything via pip method.\nIs it allow to load package to kernel as external source and then install it from local folder here at kaggle?\nhttps://pypi.org/project/tensorflow-addons/
", "rawMarkdown": "Understood.\nWhat about external packages like tensorflow-addons?\nI would like to use it but this package is not installed yet, also competition conditions don't allow me to use internet for installing anything via pip method.\nIs it allow to load package to kernel as external source and then install it from local folder here at kaggle?\nhttps://pypi.org/project/tensorflow-addons/" }, { "id": 732815, "postDate": "2020-01-30T10:03:59.100Z", "content": "You can download the package and then upload it as a dataset.\nAfter that, the package can be installed using pip in the script/notebook.
", "rawMarkdown": "You can download the package and then upload it as a dataset.\nAfter that, the package can be installed using pip in the script/notebook.", "votes": 1 }, { "id": 736501, "postDate": "2020-02-04T09:23:00.337Z", "content": "Thx, it will solve my problem.
", "rawMarkdown": "Thx, it will solve my problem." }, { "id": 738663, "postDate": "2020-02-06T20:52:31.613Z", "content": "Unfortunatelly wheel packages is not supported by kaggle kernels...
", "rawMarkdown": "Unfortunatelly wheel packages is not supported by kaggle kernels..." } ] }, { "id": 718022, "postDate": "2020-01-13T22:38:10.183Z", "content": "https://www.kaggle.com/pnussbaum/grapheme-mind-reader-panv00\nTrains the NN using GPU's\nThen this can be used for NN \"mind reading\" visualization exploration of the trained NN without GPU required. An example is here:\nhttps://www.kaggle.com/pnussbaum/grapheme-mind-reader-panv12-nogpu
", "rawMarkdown": "https://www.kaggle.com/pnussbaum/grapheme-mind-reader-panv00\nTrains the NN using GPU's\nThen this can be used for NN \"mind reading\" visualization exploration of the trained NN without GPU required. An example is here:\nhttps://www.kaggle.com/pnussbaum/grapheme-mind-reader-panv12-nogpu" }, { "id": 716638, "postDate": "2020-01-12T03:15:20.313Z", "content": "KMNIST, QMNIST, EMNIST\nfrom https://pytorch.org/docs/stable/torchvision/datasets.html\nOmniglot dataset from https://github.com/brendenlake/omniglot
", "rawMarkdown": "KMNIST, QMNIST, EMNIST\nfrom https://pytorch.org/docs/stable/torchvision/datasets.html\nOmniglot dataset from https://github.com/brendenlake/omniglot" }, { "id": 709577, "postDate": "2020-01-03T17:23:52.393Z", "content": "@sohier Is it okay to share private kaggle data as a shared link here or I have make it public? And should I share my model weight files?
", "rawMarkdown": "@sohier Is it okay to share private kaggle data as a shared link here or I have make it public? And should I share my model weight files?", "replies": [ { "id": 709596, "postDate": "2020-01-03T17:49:18.563Z", "content": "Thanks, Sudipto. An alternate link for BanglaLekha is here if anyone having problem with accessing. https://data.mendeley.com/datasets/hf6sf8zrkc/2\nIt is a 188 MB archive file.
", "rawMarkdown": "Thanks, Sudipto. An alternate link for BanglaLekha is here if anyone having problem with accessing. https://data.mendeley.com/datasets/hf6sf8zrkc/2\nIt is a 188 MB archive file." } ] } ], "comments": [ { "id": 709666, "author_name": "DrHB", "author_url": "", "post_date": "2020-01-03T19:45:27.173000", "content": "cadene pretarined models\nhttps://github.com/Cadene/pretrained-models.pytorch
torchvision.models\nhttps://pytorch.org/docs/stable/torchvision/models.html
Pytorch EfficientNet\nhttps://github.com/lukemelas/EfficientNet-PyTorch
pytorch image models\nhttps://github.com/rwightman/pytorch-image-models/tree/master/timm
\n\npytorch pretrained models\nhttps://github.com/Cadene/pretrained-models.pytorch
\n\nekush dataset\nhttps://shahariarrabby.github.io/ekush/#overview
\n\ntorchvision.models\nhttps://pytorch.org/docs/stable/torchvision/models.html
\n\nPytorch EfficientNet\nhttps://github.com/lukemelas/EfficientNet-PyTorch
", "votes": 1, "replies": [] }, { "id": 743631, "author_name": "Brian Lee", "author_url": "", "post_date": "2020-02-12T06:39:18.590000", "content": "I plan to experiemen with the following as posted by @ipythonx (since I dont see explicit links mentioned here just in case):
\n\nBanglaLekha-Isolated-Data-Set\n Ekush-Data-Set \nNumtaDB\nISI Image Data \nCMATERdb
", "votes": 1, "replies": [ { "id": 743652, "author_name": "Innat", "author_url": "", "post_date": "2020-02-12T06:53:02.603000", "content": "NumtaDB is only about bengali digit (0,1,2,..9), so you should skip this). ISI is free but need official invitation to the author to get, so bit trouble. I think if you really plan so, you should go with Ekush and BanglaLekha, more than enough.
", "votes": 6, "replies": [] }, { "id": 744525, "author_name": "Brian Lee", "author_url": "", "post_date": "2020-02-12T22:47:47.610000", "content": "I see, thank you for the information!
", "votes": 1, "replies": [] } ] }, { "id": 737174, "author_name": "Rohit Agarwal", "author_url": "", "post_date": "2020-02-05T02:35:29.800000", "content": "If I create my own labelled dataset and train the model on it, then do I have to disclose it / how do I disclose it?\nWhat if I don't disclose it - how can the organisers possibly come to know, esp. if the model is trained offline?\n@sohier
", "votes": 1, "replies": [ { "id": 738404, "author_name": "Mobassir", "author_url": "", "post_date": "2020-02-06T13:33:39.410000", "content": "i think they do hard checking only for those who finish on high,imagine you end up being in gold zone or in prize zone,then kaggle will invest more time on investigating you and your model,if they can catch you then possibly your account can be removed permanently like this : https://www.kaggle.com/c/petfinder-adoption-prediction/discussion/125436#715470 ,they take very good care for the people who finish on high,if you stay far away from gold zone in private lb then maybe they won't catch it but if they can detect that you are cheating then you are in immense trouble!
", "votes": 2, "replies": [] } ] }, { "id": 757894, "author_name": "Benlei Cui", "author_url": "", "post_date": "2020-02-27T08:09:43.723000", "content": "cadene pretarined models\nhttps://github.com/Cadene/pretrained-models.pytorch
\n\nPytorch EfficientNet\nhttps://github.com/lukemelas/EfficientNet-PyTorch
", "votes": 2, "replies": [] }, { "id": 767565, "author_name": "Guanshuo Xu", "author_url": "", "post_date": "2020-03-09T20:20:36.080000", "content": "https://github.com/MinhasKamal/BengaliDictionary
", "votes": 0, "replies": [] }, { "id": 700602, "author_name": "whale9490", "author_url": "", "post_date": "2019-12-22T09:27:41.947000", "content": "EfficientNet Keras\nhttps://github.com/qubvel/efficientnet
", "votes": 2, "replies": [] }, { "id": 775379, "author_name": "deoxy", "author_url": "", "post_date": "2020-03-16T16:13:28.933000", "content": "@vibhore94 Oh sir this is magic...hope I found it sooner :)
", "votes": 1, "replies": [] }, { "id": 775575, "author_name": "Dieter", "author_url": "", "post_date": "2020-03-16T20:25:30.547000", "content": "too late though. deadline for external data was 7 days ago
", "votes": 6, "replies": [] }, { "id": 775593, "author_name": "Camaro", "author_url": "", "post_date": "2020-03-16T20:57:46.713000", "content": "Wow, and some of them were not declared by someone before...\nBut I saw some cases that winner's solution includes extra data without disclosing it before the deadline and it seems kaggle and host doesn't check it seriously...\nBut I think they should do it more vigorously.
", "votes": 2, "replies": [] }, { "id": 775594, "author_name": "Psi", "author_url": "", "post_date": "2020-03-16T20:59:40.430000", "content": "But if he would have posted that he uses GANs, all his advantage would have been gone? Kalpurush was posted before, but not all yeah. Hard to judge.
", "votes": 3, "replies": [] }, { "id": 775628, "author_name": "Camaro", "author_url": "", "post_date": "2020-03-16T22:21:04.873000", "content": "Yes, as same as other rules it's not crystal clear. The question about which kind of data is considered as external data and which is not should be judged by host.
\n\nI'm not sure at all that the GANs are considered as external data or not, but if it's external data and if it helps to improve score, I guess that's why this rule exists. That advantage should have gone a week ago.\n(Again, I'm not judging he is violating the rule or not. It should be done by host. And also he may not use these in his final solutions.)
", "votes": 2, "replies": [] }, { "id": 775800, "author_name": "deoxy", "author_url": "", "post_date": "2020-03-17T01:10:29.067000", "content": "I'm sorry. I am new to Kaggle and did not understand the rules for publishing external data.\nCan I ask where the exact deadline could be found?
", "votes": 1, "replies": [] }, { "id": 775838, "author_name": "Camaro", "author_url": "", "post_date": "2020-03-17T01:35:02.273000", "content": "Here it is.(in C.)\nI totally understand what you're saying, it should be posted on top of this thread. Most of competitors know just because it's always same as team merge deadline.\nIn my honest opinion it's pretty minor thing compared to what you achieved. Congrats and you definitely deserve it!\n
https://www.kaggle.com/c/bengaliai-cv19/rules
", "votes": 6, "replies": [] } ] }, { "id": 767580, "author_name": "Md Yasin Kabir", "author_url": "", "post_date": "2020-03-09T20:41:12.393000", "content": "I have created a set of synthetic characters using the following Bengali fonts:\nEkushey_Durga-Normal.ttf\nEkushey_Sharifa-Normal.ttf\nEkushey_Puja-Normal.ttf\nGalada-Regular.ttf\nEkushey_Punarbhaba-Normal.ttf\nBenSenHandwriting.ttf\nEkushey_Godhuli-Normal.ttf\nEkushey_Sumit-Normal.ttf\nSiyamrupali.ttf\nAtma-Bold.ttf\nBaloo_Da-Regular.ttf\nEkushey_Mohua-Normal.ttf\nKalpurush-Regular.ttf
", "votes": 0, "replies": [ { "id": 772346, "author_name": "Nicholas Lyu", "author_url": "", "post_date": "2020-03-15T11:40:26.513000", "content": "Also created a set of synthetic characters using the following fonts:\nAmarBangl, Bensenhandwriting, CHITMI, ChitraMJ, Durga, JugantorMJ, kalpurush, KeertankhuIaM, KumarkhaliMJ, Kumarkhali, Mohua, Nikosh, PANDM, Puja, PunarBhaba, Sharifa, Siyamrupali, SolaimanLi, SUTOM, SutonnyOM, TEESM
", "votes": 0, "replies": [] }, { "id": 775534, "author_name": "DrHB", "author_url": "", "post_date": "2020-03-16T19:20:11.463000", "content": "I will just add more fonts:
\n\nhttp://www.freebanglafont.com/download.php?id=649\nhttp://www.freebanglafont.com/download.php?id=762\nhttps://www.freeforfonts.com/kohinoor-bangla-font-family/\nhttp://www.freebanglafont.com/download.php?id=765
\n\nhttp://www.freebanglafont.com/download.php?id=659\nhttps://www.stylemyname.com/Chit-Mi\nhttp://www.freebanglafont.com/download.php?id=708\nhttp://www.freebanglafont.com/download.php?id=724\nhttp://www.freebanglafont.com/download.php?id=769\nhttps://blogfonts.com/ekushey-punarbhaba.font\nhttp://www.freebanglafont.com/download.php?id=658\nhttp://www.freebanglafont.com/download.php?id=700\nhttp://www.freebanglafont.com/download.php?id=690\nhttp://www.freebanglafont.com/download.php?id=763\nhttp://www.freebanglafont.com/download.php?id=708\n.TTF\">https://github.com/torifat/bangla-font/blob/master/src/fonts/SUTOM_.TTF\nhttp://www.freebanglafont.com/download.php?id=651
", "votes": 0, "replies": [] }, { "id": 775542, "author_name": "GreatGameDota", "author_url": "", "post_date": "2020-03-16T19:35:37.677000", "content": "Looks like you guys Learned the Magic! Grats!
", "votes": 1, "replies": [] }, { "id": 775572, "author_name": "", "author_url": "", "post_date": "2020-03-16T20:21:53.463000", "content": "", "votes": 0, "replies": [] }, { "id": 775573, "author_name": "See--", "author_url": "", "post_date": "2020-03-16T20:23:39.123000", "content": "… but you are not allowed to use them in your solution. The deadline was 1 week ago: March 9, 2020.
", "votes": 5, "replies": [] } ] }, { "id": 775519, "author_name": "pkugoodspeed", "author_url": "", "post_date": "2020-03-16T19:03:20.167000", "content": "https://www.kaggle.com/pkugoodspeed/bengali-competition-checkpoints\nThanks
", "votes": 0, "replies": [] }, { "id": 775464, "author_name": "madmax0404", "author_url": "", "post_date": "2020-03-16T17:48:42.803000", "content": "https://keras.io/applications/#densenet
", "votes": 0, "replies": [] }, { "id": 775141, "author_name": "Philip Popien", "author_url": "", "post_date": "2020-03-16T10:59:59.630000", "content": "https://github.com/rwightman/gen-efficientnet-pytorch
", "votes": 0, "replies": [] }, { "id": 774937, "author_name": "yhfu", "author_url": "", "post_date": "2020-03-16T04:43:38.647000", "content": "https://github.com/Callidior/keras-applications/releases/download/efficientnet\nhttps://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b4_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5\n https://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b3_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5\n https://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b5_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5
", "votes": 0, "replies": [] }, { "id": 771819, "author_name": "Matt Yates", "author_url": "", "post_date": "2020-03-14T16:59:33.120000", "content": "https://www.kaggle.com/yeayates21/bengaliaitrainedmodelv4
", "votes": 0, "replies": [] }, { "id": 767524, "author_name": "Max Söderman", "author_url": "", "post_date": "2020-03-09T18:41:58.873000", "content": "https://www.kaggle.com/lucky227/ooooo3\nhttps://www.kaggle.com/rsmits/kerasefficientnetb3\nhttps://www.kaggle.com/h030162/se-resnext50-32x4d-fold2\nhttps://www.kaggle.com/h030162/se-resnext50-32x4d-fold3
", "votes": 0, "replies": [] }, { "id": 767427, "author_name": "momi64", "author_url": "", "post_date": "2020-03-09T16:07:00.643000", "content": "https://github.com/adobe/antialiased-cnns
", "votes": 0, "replies": [] }, { "id": 767309, "author_name": "Arthur Stsepanenka", "author_url": "", "post_date": "2020-03-09T13:18:17.530000", "content": "http://places2.csail.mit.edu/index.html
", "votes": 0, "replies": [] }, { "id": 767290, "author_name": "Mohammad Azam Khan", "author_url": "", "post_date": "2020-03-09T12:49:11.197000", "content": "cadene pre-trained models: https://github.com/Cadene/pretrained-models.pytorch\ntorchvision.models: https://pytorch.org/docs/stable/torchvision/models.html
", "votes": 0, "replies": [] }, { "id": 767041, "author_name": "Shai", "author_url": "", "post_date": "2020-03-09T04:32:23.433000", "content": "I may use: https://github.com/microsoft/computervision-recipes
", "votes": 0, "replies": [] }, { "id": 766892, "author_name": "Appian", "author_url": "", "post_date": "2020-03-08T22:53:47.007000", "content": "https://github.com/Cadene/pretrained-models.pytorch\nhttps://github.com/lukemelas/EfficientNet-PyTorch\nhttps://shahariarrabby.github.io/ekush/\nhttp://www.openslr.org/37/\nhttp://www.openslr.org/53/
", "votes": 0, "replies": [] }, { "id": 766389, "author_name": "Josef Slavicek", "author_url": "", "post_date": "2020-03-08T04:40:03.673000", "content": "maybe I will try to somehow use this (it's chinese):
\n\nhttp://www.nlpr.ia.ac.cn/databases/handwriting/Home.html
\n\nhttp://www.nlpr.ia.ac.cn/databases/handwriting/Download.html
\n\nhttp://www.nlpr.ia.ac.cn/databases/Download/feature_data/1.0train-gb1.rar
\n\nhttp://www.nlpr.ia.ac.cn/databases/Download/feature_data/1.0test-gb1.rar
", "votes": 0, "replies": [] }, { "id": 766145, "author_name": "João Guilherme", "author_url": "", "post_date": "2020-03-07T18:22:57.073000", "content": "https://github.com/OverLordGoldDragon/keras-adamw - AdamW for Keras\nhttps://github.com/albu/albumentations - Albumentations library for data augmentation
", "votes": 0, "replies": [] }, { "id": 765860, "author_name": "Ruslan Baikulov", "author_url": "", "post_date": "2020-03-07T08:30:05.017000", "content": "PyTorch Image Models\nhttps://github.com/rwightman/pytorch-image-models
", "votes": 0, "replies": [] }, { "id": 765584, "author_name": "Helen", "author_url": "", "post_date": "2020-03-06T20:34:24.887000", "content": "cadene pretarined models\nhttps://github.com/Cadene/pretrained-models.pytorch
", "votes": 0, "replies": [] }, { "id": 761313, "author_name": "Juhyeong", "author_url": "", "post_date": "2020-03-02T11:34:51.427000", "content": "https://pypi.org/project/efficientnet/
", "votes": 0, "replies": [] }, { "id": 760527, "author_name": "Sachin Prabhu", "author_url": "", "post_date": "2020-03-01T12:17:10.867000", "content": "https://github.com/Cadene/pretrained-models.pytorch\nhttps://github.com/lukemelas/EfficientNet-PyTorch\nhttps://github.com/osmr/imgclsmob\nhttps://github.com/qubvel/segmentation_models.pytorch
", "votes": 0, "replies": [] }, { "id": 760345, "author_name": "عثمان", "author_url": "", "post_date": "2020-03-01T06:37:31.903000", "content": "EfficientNet models from Tensorflow repo https://github.com/tensorflow/tpu/tree/69a06b6a11ff9ba8835b27348f66f33864af93fe/models/official/efficientnet https://github.com/Kyubyong/wordvectors\nhttps://pypi.org/project/grapheme/
", "votes": 0, "replies": [] }, { "id": 759979, "author_name": "Shayekh Islam", "author_url": "", "post_date": "2020-02-29T17:24:59.090000", "content": "", "votes": 0, "replies": [] }, { "id": 759879, "author_name": "None", "author_url": "", "post_date": "2020-02-29T15:17:08.263000", "content": "", "votes": 0, "replies": [] }, { "id": 759852, "author_name": "", "author_url": "", "post_date": "2020-02-29T14:47:33.987000", "content": "", "votes": 0, "replies": [] }, { "id": 759847, "author_name": "", "author_url": "", "post_date": "2020-02-29T14:42:07.583000", "content": "", "votes": 0, "replies": [] }, { "id": 759304, "author_name": "", "author_url": "", "post_date": "2020-02-28T21:40:21.327000", "content": "", "votes": 0, "replies": [] }, { "id": 755231, "author_name": "", "author_url": "", "post_date": "2020-02-24T15:25:04.997000", "content": "", "votes": 0, "replies": [] }, { "id": 749097, "author_name": "", "author_url": "", "post_date": "2020-02-18T09:42:15.683000", "content": "", "votes": 0, "replies": [] }, { "id": 748570, "author_name": "", "author_url": "", "post_date": "2020-02-17T17:13:28.663000", "content": "", "votes": 0, "replies": [] }, { "id": 746389, "author_name": "", "author_url": "", "post_date": "2020-02-14T23:43:36.660000", "content": "", "votes": 0, "replies": [ { "id": 748545, "author_name": "", "author_url": "", "post_date": "2020-02-17T16:40:00.617000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 740745, "author_name": "", "author_url": "", "post_date": "2020-02-09T18:14:27.070000", "content": "", "votes": 0, "replies": [] }, { "id": 740345, "author_name": "", "author_url": "", "post_date": "2020-02-09T09:49:13.297000", "content": "", "votes": 0, "replies": [] }, { "id": 740059, "author_name": "", "author_url": "", "post_date": "2020-02-08T19:25:16.100000", "content": "", "votes": 0, "replies": [] }, { "id": 734750, "author_name": "", "author_url": "", "post_date": "2020-02-02T00:31:51.103000", "content": "", "votes": 0, "replies": [ { "id": 735796, "author_name": "", "author_url": "", "post_date": "2020-02-03T13:20:49.177000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 725713, "author_name": "", "author_url": "", "post_date": "2020-01-22T12:27:11.790000", "content": "", "votes": 0, "replies": [ { "id": 725981, "author_name": "", "author_url": "", "post_date": "2020-01-22T17:27:14.430000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 722356, "author_name": "", "author_url": "", "post_date": "2020-01-18T13:06:07.193000", "content": "", "votes": 0, "replies": [ { "id": 725982, "author_name": "", "author_url": "", "post_date": "2020-01-22T17:27:47.057000", "content": "", "votes": 0, "replies": [] }, { "id": 727578, "author_name": "", "author_url": "", "post_date": "2020-01-23T20:53:43.533000", "content": "", "votes": 0, "replies": [] }, { "id": 732815, "author_name": "", "author_url": "", "post_date": "2020-01-30T10:03:59.100000", "content": "", "votes": 1, "replies": [] }, { "id": 736501, "author_name": "", "author_url": "", "post_date": "2020-02-04T09:23:00.337000", "content": "", "votes": 0, "replies": [] }, { "id": 738663, "author_name": "", "author_url": "", "post_date": "2020-02-06T20:52:31.613000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 718022, "author_name": "", "author_url": "", "post_date": "2020-01-13T22:38:10.183000", "content": "", "votes": 0, "replies": [] }, { "id": 716638, "author_name": "", "author_url": "", "post_date": "2020-01-12T03:15:20.313000", "content": "", "votes": 0, "replies": [] }, { "id": 709577, "author_name": "", "author_url": "", "post_date": "2020-01-03T17:23:52.393000", "content": "", "votes": 0, "replies": [ { "id": 709596, "author_name": "", "author_url": "", "post_date": "2020-01-03T17:49:18.563000", "content": "", "votes": 2, "replies": [] } ] }, { "id": 767083, "author_name": "", "author_url": "", "post_date": "2020-03-09T06:17:52.803000", "content": "", "votes": 0, "replies": [] }, { "id": 766155, "author_name": "", "author_url": "", "post_date": "2020-03-07T18:42:07.540000", "content": "", "votes": 0, "replies": [] }, { "id": 764002, "author_name": "", "author_url": "", "post_date": "2020-03-05T04:13:50.553000", "content": "", "votes": 0, "replies": [] }, { "id": 753925, "author_name": "", "author_url": "", "post_date": "2020-02-22T20:39:40.370000", "content": "", "votes": 0, "replies": [] }, { "id": 750713, "author_name": "", "author_url": "", "post_date": "2020-02-19T16:25:23.937000", "content": "", "votes": 0, "replies": [] }, { "id": 730162, "author_name": "", "author_url": "", "post_date": "2020-01-27T07:01:15.587000", "content": "", "votes": 3, "replies": [ { "id": 738032, "author_name": "", "author_url": "", "post_date": "2020-02-06T04:34:42.470000", "content": "", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "698947": "Post links to your external data sources here before the deadline specified in the rules.", "709666": "`cadene pretarined models`\nhttps://github.com/Cadene/pretrained-models.pytorch\n\n`torchvision.models`\nhttps://pytorch.org/docs/stable/torchvision/models.html\n\n`Pytorch EfficientNet`\nhttps://github.com/lukemelas/EfficientNet-PyTorch", "766523": "pytorch image models\nhttps://github.com/rwightman/pytorch-image-models/tree/master/timm\n\npytorch pretrained models\nhttps://github.com/Cadene/pretrained-models.pytorch\n\nekush dataset\nhttps://shahariarrabby.github.io/ekush/#overview\n\ntorchvision.models\nhttps://pytorch.org/docs/stable/torchvision/models.html\n\nPytorch EfficientNet\nhttps://github.com/lukemelas/EfficientNet-PyTorch\n", "743631": "I plan to experiemen with the following as posted by @ipythonx (since I dont see explicit links mentioned here just in case): \n\n[ BanglaLekha-Isolated-Data-Set](https://data.mendeley.com/datasets/hf6sf8zrkc/2?__cf_chl_captcha_tk__=0eef3f155bb8cdbd4318462de9ce43b06e8b36fe-1576926645-0-ARwIJOlc0pqWADHpP7QJNSnlLmWPdU95TWcw8RABhCTk-MHuKadBfXUxQednw2omNlWNN2YFffaQjm6HFtzYhbKIaX6ZXIydnz6nMJUQL1p3MoCjLaD8C1Y7P2F33-MAWE3Eo1NstIjhrEzpMBoz4nmc0z8Fx1uftg5rDt5YCNCiWsBY4VcaijT_NXaReFueYu3CBgk-he-9c_Xr7FhsPoQCpg8ThfHCmELrGY8f5EKmi5eFhBEWP6-1JpByjl4Z4Qr5-3zWmjn0uBsoPF2DKD6SE10QnF1hcUP5rHbSfV4Iv8mLUYoTLLeArXclMmeNEQuWKo_AgS5fAzLFuFCf_DI)\n [Ekush-Data-Set](https://shahariarrabby.github.io/ekush/#overview) \n[NumtaDB](https://www.kaggle.com/BengaliAI/numta)\n[ISI Image Data](https://www.isical.ac.in/~ujjwal/download/BanglaNumeral.html) \n[CMATERdb](https://code.google.com/archive/p/cmaterdb/)\n", "737174": "If I create my own labelled dataset and train the model on it, then do I have to disclose it / how do I disclose it?\nWhat if I don't disclose it - how can the organisers possibly come to know, esp. if the model is trained offline?\n@sohier ", "757894": "cadene pretarined models\nhttps://github.com/Cadene/pretrained-models.pytorch\n\nPytorch EfficientNet\nhttps://github.com/lukemelas/EfficientNet-PyTorch", "767565": "https://github.com/MinhasKamal/BengaliDictionary", "700602": "EfficientNet Keras\nhttps://github.com/qubvel/efficientnet", "775379": "\n* [https://github.com/lukemelas/EfficientNet-PyTorch](https://github.com/lukemelas/EfficientNet-PyTorch)\n* [https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix](https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix)\n* [https://github.com/DeepVoltaire/AutoAugment](https://github.com/DeepVoltaire/AutoAugment)\n* https://www.omicronlab.com/download/fonts/kalpurush.ttf\n* https://www.omicronlab.com/download/fonts/NikoshLightBan.ttf\n* [http://www.freebanglafont.com/download.php?id=659](http://www.freebanglafont.com/download.php?id=659)\n\n", "767580": "I have created a set of synthetic characters using the following Bengali fonts:\nEkushey_Durga-Normal.ttf\nEkushey_Sharifa-Normal.ttf\nEkushey_Puja-Normal.ttf\nGalada-Regular.ttf\nEkushey_Punarbhaba-Normal.ttf\nBenSenHandwriting.ttf\nEkushey_Godhuli-Normal.ttf\nEkushey_Sumit-Normal.ttf\nSiyamrupali.ttf\nAtma-Bold.ttf\nBaloo_Da-Regular.ttf\nEkushey_Mohua-Normal.ttf\nKalpurush-Regular.ttf", "775519": "https://www.kaggle.com/pkugoodspeed/bengali-competition-checkpoints\nThanks", "775464": "https://keras.io/applications/#densenet", "775141": "https://github.com/rwightman/gen-efficientnet-pytorch", "774937": "https://github.com/Callidior/keras-applications/releases/download/efficientnet\nhttps://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b4_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5\n https://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b3_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5\n https://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b5_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5", "771819": "https://www.kaggle.com/yeayates21/bengaliaitrainedmodelv4", "767524": "https://www.kaggle.com/lucky227/ooooo3\nhttps://www.kaggle.com/rsmits/kerasefficientnetb3\nhttps://www.kaggle.com/h030162/se-resnext50-32x4d-fold2\nhttps://www.kaggle.com/h030162/se-resnext50-32x4d-fold3", "767427": "https://github.com/adobe/antialiased-cnns", "767309": "http://places2.csail.mit.edu/index.html", "767290": "cadene pre-trained models: https://github.com/Cadene/pretrained-models.pytorch\ntorchvision.models: https://pytorch.org/docs/stable/torchvision/models.html", "767041": "I may use: https://github.com/microsoft/computervision-recipes", "766892": "https://github.com/Cadene/pretrained-models.pytorch\nhttps://github.com/lukemelas/EfficientNet-PyTorch\nhttps://shahariarrabby.github.io/ekush/\nhttp://www.openslr.org/37/\nhttp://www.openslr.org/53/", "766389": "maybe I will try to somehow use this (it's chinese):\n\nhttp://www.nlpr.ia.ac.cn/databases/handwriting/Home.html\n\nhttp://www.nlpr.ia.ac.cn/databases/handwriting/Download.html\n\nhttp://www.nlpr.ia.ac.cn/databases/Download/feature_data/1.0train-gb1.rar\n\nhttp://www.nlpr.ia.ac.cn/databases/Download/feature_data/1.0test-gb1.rar", "766145": "https://github.com/OverLordGoldDragon/keras-adamw - AdamW for Keras\nhttps://github.com/albu/albumentations - Albumentations library for data augmentation", "765860": "PyTorch Image Models\nhttps://github.com/rwightman/pytorch-image-models", "765584": "cadene pretarined models\nhttps://github.com/Cadene/pretrained-models.pytorch", "761313": "https://pypi.org/project/efficientnet/", "760527": "https://github.com/Cadene/pretrained-models.pytorch\nhttps://github.com/lukemelas/EfficientNet-PyTorch\nhttps://github.com/osmr/imgclsmob\nhttps://github.com/qubvel/segmentation_models.pytorch", "760345": "EfficientNet models from Tensorflow repo https://github.com/tensorflow/tpu/tree/69a06b6a11ff9ba8835b27348f66f33864af93fe/models/official/efficientnet https://github.com/Kyubyong/wordvectors\nhttps://pypi.org/project/grapheme/", "759979": "https://pypi.org/project/efficientnet/", "759879": "https://github.com/aitorzip/PyTorch-CycleGAN", "759852": "Pretrained models: https://github.com/osmr/imgclsmob", "759847": "keras-efficientnets\nhttps://pypi.org/project/keras-efficientnets/", "759304": "https://github.com/tensorflow/models/\nhttps://github.com/qubvel/classification_models/tree/master/classification_models/models\nhttps://www.kaggle.com/rsmits/kerasefficientnetb3\nhttps://github.com/qubvel/efficientnet", "755231": "Should we disclosure private scripts and libraries we use as well? I have a couple of packages I use as helpers. (Some of them are available on public GitHub, while others are not). Should I list them here as well? None of them includes any pre-trained weights or training data.", "749097": "Is it possible to use generated dataset? Symbols generated by draw it on context (like [here](https://www.kaggle.com/kaushal2896/bengali-graphemes-starter-eda-multi-output-cnn))", "748570": "I don't see any explicit mention of the transfer learning models available via keras/tensorflow, so I figured I'd post them here just in case they'd be of help to anyone (and so I can use them 😃).\n\nhttps://www.tensorflow.org/api_docs/python/tf/keras/applications", "746389": "If someone has already has posted an external data set that I am using, am I required to also post it as a disclosure?\n\n(I am using `torchvision.models` found here https://pytorch.org/docs/stable/torchvision/models.html)", "740745": "Are we allowed to install new packages in the kaggle env ? \n\nLike in : https://www.kaggle.com/c/severstal-steel-defect-detection/discussion/109679 ? \n\nIt will be a lot easier than to copy past code into a single script.\n", "740345": "keras pretrained from https://github.com/qubvel/classification_models", "740059": "Using pretraining on https://www.kaggle.com/rishianand/devanagari-character-set", "734750": "Is that fair that I train the model using a huge amount of resources, then save and upload the weights (.h5) to the kernel and make inference only?", "725713": "if I train a generative model offline for data augmentation and use the augmented data in a kaggle kernel, what do i need to publish here? \n", "722356": "As I assumed this competition allow me to use pretrained models and just make predictions using test data via kernel here at Kaggle for submitting. Can you tell my what is forbidden? Should I make pretrained weights public or something something else? I'm newbie in this thread, don't get me wrong please.", "718022": "https://www.kaggle.com/pnussbaum/grapheme-mind-reader-panv00\nTrains the NN using GPU's\nThen this can be used for NN \"mind reading\" visualization exploration of the trained NN without GPU required. An example is here:\nhttps://www.kaggle.com/pnussbaum/grapheme-mind-reader-panv12-nogpu", "716638": "KMNIST, QMNIST, EMNIST\nfrom https://pytorch.org/docs/stable/torchvision/datasets.html\nOmniglot dataset from https://github.com/brendenlake/omniglot", "709577": "@sohier Is it okay to share private kaggle data as a shared link here or I have make it public? And should I share my model weight files?", "767083": "", "766155": "", "764002": "", "753925": "", "750713": "", "730162": "" } }