Use this thread to post sources of external data that you use as part of your approach.
", "messages": [ { "id": 210932, "postDate": "2017-08-07T16:47:13.753Z", "content": "Use this thread to post sources of external data that you use as part of your approach.
", "rawMarkdown": "Use this thread to post sources of external data that you use as part of your approach.", "votes": 3 }, { "id": 245833, "postDate": "2017-11-19T21:02:37.403Z", "content": "If someone else has already posted external data that we use, is it necessary to disclose our own use of it as well? For example, I thought I'd try using some pretrained weights with ResNet to see if I get better performance than what I've put together: if it scores better, do I need to post my own disclosure of that, or is staying quiet because somebody else did fine?
\n\nThanks :)
", "rawMarkdown": "If someone else has already posted external data that we use, is it necessary to disclose our own use of it as well? For example, I thought I'd try using some pretrained weights with ResNet to see if I get better performance than what I've put together: if it scores better, do I need to post my own disclosure of that, or is staying quiet because somebody else did fine?\n\nThanks :)", "votes": 1, "replies": [ { "id": 246177, "postDate": "2017-11-20T16:37:25.337Z", "content": "You don't have to repost if someone else has already posted it.
", "rawMarkdown": "You don't have to repost if someone else has already posted it." } ] }, { "id": 216491, "postDate": "2017-08-26T01:21:40.973Z", "content": "Are we allowed to use external data in this competition? From reading the rules, it says
\n\n\n\n\nEXTERNAL DATA\n Unless otherwise expressly stated on the Competition Website, Participants must not use data other than the Data to develop and test their models and Submissions. Competition Sponsor reserves the right in its sole discretion to disqualify any Participant who Competition Sponsor discovers has undertaken or attempted to undertake the use of data other than the Data, or who uses the Data other than as permitted according to the Competition Website and in these Competition Rules, in the course of the Competition.
\n
but I don't see where it's \"expressly stated\" that we can use external data.
", "rawMarkdown": "Are we allowed to use external data in this competition? From reading the rules, it says\n\n> EXTERNAL DATA\nUnless otherwise expressly stated on the Competition Website, Participants must not use data other than the Data to develop and test their models and Submissions. Competition Sponsor reserves the right in its sole discretion to disqualify any Participant who Competition Sponsor discovers has undertaken or attempted to undertake the use of data other than the Data, or who uses the Data other than as permitted according to the Competition Website and in these Competition Rules, in the course of the Competition.\n\nbut I don't see where it's \"expressly stated\" that we can use external data.", "votes": 2, "replies": [ { "id": 216492, "postDate": "2017-08-26T01:33:10.160Z", "content": "(oops, nevermind -- this is explicitly allowed earlier in the rules)
", "rawMarkdown": "(oops, nevermind -- this is explicitly allowed earlier in the rules)", "votes": 1 } ] }, { "id": 253430, "postDate": "2017-12-04T23:53:05.240Z", "content": "For the final model, I may use one or more of the following, data and/or pre-trained models:
\n\nPre-trained models and external datasets used zoo. The copyrights belong to the original authors. (I'm using only some of them of course (!) wherever allowed, but I thought the extended listing may be of use to others. Sorry for the length!):
\n\nExternal datasets (note: some also include published pre-trained benchmark models):
\n\nhttps://www.cs.toronto.edu/~kriz/cifar.html
\n\nhttps://deepmind.com/research/open-source/open-source-datasets/kinetics/
\n\nhttps://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI
\n\nhttp://cs.stanford.edu/people/karpathy/deepvideo/
\n\nhttp://human-pose.mpi-inf.mpg.de/
\n\nhttp://humanshape.mpi-inf.mpg.de/
\n\nhttp://homepages.inf.ed.ac.uk/rbf/CVonline/Imagedbase.htm
\n\nhttp://yann.lecun.com/exdb/mnist/
\n\n\n\nPre-trained models:
\n\nhttp://data.dmlc.ml/mxnet/models/
\n\nhttp://vlg.cs.dartmouth.edu/c3d/
\n\nhttp://humanshape.mpi-inf.mpg.de/#results
\n\nhttps://github.com/tensorflow/models/
\n\nhttps://mxnet.incubator.apache.org/model_zoo/
\n\nhttps://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/
\n\nhttps://github.com/dmlc/mxnet-model-gallery/
\n\nhttps://github.com/rai-project/mxnet/tree/master/builtin_models/
\n\nhttps://github.com/deepmind/kinetics-i3d/
\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/
\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/rcnn/
\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/image-classification/
\n\nhttps://github.com/tensorflow/models/tree/master/research/slim/
\n\nhttps://github.com/KeyKy/model-zoo/
\n\nhttps://github.com/xingyizhou/pose-hg-3d/
\n\nhttps://github.com/liuzhuang13/DenseNet/
\n\nhttps://github.com/SimJeg/FC-DenseNet/
\n\nhttps://github.com/ShichenLiu/CondenseNet/
\n\nhttps://github.com/msracver/Deformable-ConvNets/
\n\nhttps://github.com/szq0214/DSOD/
\n\nhttps://github.com/szagoruyko/functional-zoo/
\n\nhttps://github.com/JaggerYoung/C3D-mxnet/
\n\nhttps://github.com/soeaver/caffe-model/
\n\nhttps://github.com/rbgirshick/py-faster-rcnn/
\n\nhttps://github.com/YuwenXiong/py-R-FCN/
\n\nhttps://github.com/juliandewit/kaggle_ndsb2017/
\n\nhttps://github.com/dhammack/DSB2017/
\n\n\n\nhttps://github.com/bearpaw/PyraNet/
\n\nhttps://github.com/anewell/pose-hg-train/
", "rawMarkdown": "Pre-trained models and external datasets used zoo. The copyrights belong to the original authors. (I'm using only some of them of course (!) wherever allowed, but I thought the extended listing may be of use to others. Sorry for the length!):\n\n\n----------\n\nExternal datasets (note: some also include published pre-trained benchmark models):\n\nhttps://www.cs.toronto.edu/~kriz/cifar.html\n\nhttps://deepmind.com/research/open-source/open-source-datasets/kinetics/\n\nhttps://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI\n\nhttp://cs.stanford.edu/people/karpathy/deepvideo/\n\nhttp://human-pose.mpi-inf.mpg.de/\n\nhttp://humanshape.mpi-inf.mpg.de/\n\nhttp://homepages.inf.ed.ac.uk/rbf/CVonline/Imagedbase.htm\n\nhttp://yann.lecun.com/exdb/mnist/\n\nhttp://www.image-net.org\n\n----------\n\nPre-trained models:\n\nhttp://data.dmlc.ml/mxnet/models/\n\nhttp://vlg.cs.dartmouth.edu/c3d/\n\nhttp://humanshape.mpi-inf.mpg.de/#results\n\nhttps://github.com/tensorflow/models/\n\nhttps://mxnet.incubator.apache.org/model_zoo/\n\nhttps://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/\n\nhttps://github.com/dmlc/mxnet-model-gallery/\n\nhttps://github.com/rai-project/mxnet/tree/master/builtin_models/\n\nhttps://github.com/deepmind/kinetics-i3d/\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/rcnn/\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/image-classification/\n\nhttps://github.com/tensorflow/models/tree/master/research/slim/\n\nhttps://github.com/KeyKy/model-zoo/\n\nhttps://github.com/xingyizhou/pose-hg-3d/\n\nhttps://github.com/liuzhuang13/DenseNet/\n\nhttps://github.com/SimJeg/FC-DenseNet/\n\nhttps://github.com/ShichenLiu/CondenseNet/\n\nhttps://github.com/msracver/Deformable-ConvNets/\n\nhttps://github.com/szq0214/DSOD/\n\nhttps://github.com/szagoruyko/functional-zoo/\n\nhttps://github.com/JaggerYoung/C3D-mxnet/\n\nhttps://github.com/soeaver/caffe-model/\n\nhttps://github.com/rbgirshick/py-faster-rcnn/\n\nhttps://github.com/YuwenXiong/py-R-FCN/\n\nhttps://github.com/juliandewit/kaggle_ndsb2017/\n\nhttps://github.com/dhammack/DSB2017/\n\nhttps://github.com/handong1587/handong1587.github.io/blob/master/_posts/deep_learning/2015-10-09-segmentation.md\n\nhttps://github.com/bearpaw/PyraNet/\n\nhttps://github.com/anewell/pose-hg-train/", "votes": -1, "replies": [ { "id": 254329, "postDate": "2017-12-06T18:15:48.383Z", "content": "> USP wrote\n> \n> > Pre-trained models and external datasets used zoo. The copyrights belong to the original authors. (I'm using only some of them of course (!) wherever allowed, but I thought the extended listing may be of use to others. Sorry for the length!):\n> \n> \n> ----------\n> \n> External datasets (note: some also include published pre-trained benchmark models):\n> \n> https://www.cs.toronto.edu/~kriz/cifar.html\n> \n> https://deepmind.com/research/open-source/open-source-datasets/kinetics/\n> \n> https://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI\n> \n> http://cs.stanford.edu/people/karpathy/deepvideo/\n> \n> http://human-pose.mpi-inf.mpg.de/\n> \n> http://humanshape.mpi-inf.mpg.de/\n> \n> http://homepages.inf.ed.ac.uk/rbf/CVonline/Imagedbase.htm\n> \n> http://yann.lecun.com/exdb/mnist/\n> \n> http://www.image-net.org\n> \n> ----------\n> \n> Pre-trained models:\n> \n> http://data.dmlc.ml/mxnet/models/\n> \n> http://vlg.cs.dartmouth.edu/c3d/\n> \n> http://humanshape.mpi-inf.mpg.de/#results\n> \n> https://github.com/tensorflow/models/\n> \n> https://mxnet.incubator.apache.org/model_zoo/\n> \n> https://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/\n> \n> https://github.com/dmlc/mxnet-model-gallery/\n> \n> https://github.com/rai-project/mxnet/tree/master/builtin_models/\n> \n> https://github.com/deepmind/kinetics-i3d/\n> \n> https://github.com/apache/incubator-mxnet/tree/master/example/\n> \n> https://github.com/apache/incubator-mxnet/tree/master/example/rcnn/\n> \n> https://github.com/apache/incubator-mxnet/tree/master/example/image-classification/\n> \n> https://github.com/tensorflow/models/tree/master/research/slim/\n> \n> https://github.com/KeyKy/model-zoo/\n> \n> https://github.com/xingyizhou/pose-hg-3d/\n> \n> https://github.com/liuzhuang13/DenseNet/\n> \n> https://github.com/SimJeg/FC-DenseNet/\n> \n> https://github.com/ShichenLiu/CondenseNet/\n> \n> https://github.com/msracver/Deformable-ConvNets/\n> \n> https://github.com/szq0214/DSOD/\n> \n> https://github.com/szagoruyko/functional-zoo/\n> \n> https://github.com/JaggerYoung/C3D-mxnet/\n> \n> https://github.com/soeaver/caffe-model/\n> \n> https://github.com/rbgirshick/py-faster-rcnn/\n> \n> https://github.com/YuwenXiong/py-R-FCN/\n> \n> https://github.com/juliandewit/kaggle_ndsb2017/\n> \n> https://github.com/dhammack/DSB2017/\n> \n> https://github.com/handong1587/handong1587.github.io/blob/master/_posts/deep_learning/2015-10-09-segmentation.md\n> \n> https://github.com/bearpaw/PyraNet/\n> \n> https://github.com/anewell/pose-hg-train/
\n\nFYI, your list includes dhammack's Github repository. He's also a contestant; that decision is up to Kaggle, but technically, you're both disqualified for sharing code outside of teams I guess. That's at their discretion though, of course.
\n\n... Sorry, but if I'm going to lose, I'm only going to lose to people who play by the rules :(
", "rawMarkdown": "\n> **USP wrote**\n> \n> > Pre-trained models and external datasets used zoo. The copyrights belong to the original authors. (I'm using only some of them of course (!) wherever allowed, but I thought the extended listing may be of use to others. Sorry for the length!):\n> \n> \n> ----------\n> \n> External datasets (note: some also include published pre-trained benchmark models):\n> \n> https://www.cs.toronto.edu/~kriz/cifar.html\n> \n> https://deepmind.com/research/open-source/open-source-datasets/kinetics/\n> \n> https://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI\n> \n> http://cs.stanford.edu/people/karpathy/deepvideo/\n> \n> http://human-pose.mpi-inf.mpg.de/\n> \n> http://humanshape.mpi-inf.mpg.de/\n> \n> http://homepages.inf.ed.ac.uk/rbf/CVonline/Imagedbase.htm\n> \n> http://yann.lecun.com/exdb/mnist/\n> \n> http://www.image-net.org\n> \n> ----------\n> \n> Pre-trained models:\n> \n> http://data.dmlc.ml/mxnet/models/\n> \n> http://vlg.cs.dartmouth.edu/c3d/\n> \n> http://humanshape.mpi-inf.mpg.de/#results\n> \n> https://github.com/tensorflow/models/\n> \n> https://mxnet.incubator.apache.org/model_zoo/\n> \n> https://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/\n> \n> https://github.com/dmlc/mxnet-model-gallery/\n> \n> https://github.com/rai-project/mxnet/tree/master/builtin_models/\n> \n> https://github.com/deepmind/kinetics-i3d/\n> \n> https://github.com/apache/incubator-mxnet/tree/master/example/\n> \n> https://github.com/apache/incubator-mxnet/tree/master/example/rcnn/\n> \n> https://github.com/apache/incubator-mxnet/tree/master/example/image-classification/\n> \n> https://github.com/tensorflow/models/tree/master/research/slim/\n> \n> https://github.com/KeyKy/model-zoo/\n> \n> https://github.com/xingyizhou/pose-hg-3d/\n> \n> https://github.com/liuzhuang13/DenseNet/\n> \n> https://github.com/SimJeg/FC-DenseNet/\n> \n> https://github.com/ShichenLiu/CondenseNet/\n> \n> https://github.com/msracver/Deformable-ConvNets/\n> \n> https://github.com/szq0214/DSOD/\n> \n> https://github.com/szagoruyko/functional-zoo/\n> \n> https://github.com/JaggerYoung/C3D-mxnet/\n> \n> https://github.com/soeaver/caffe-model/\n> \n> https://github.com/rbgirshick/py-faster-rcnn/\n> \n> https://github.com/YuwenXiong/py-R-FCN/\n> \n> https://github.com/juliandewit/kaggle_ndsb2017/\n> \n> https://github.com/dhammack/DSB2017/\n> \n> https://github.com/handong1587/handong1587.github.io/blob/master/_posts/deep_learning/2015-10-09-segmentation.md\n> \n> https://github.com/bearpaw/PyraNet/\n> \n> https://github.com/anewell/pose-hg-train/\n\nFYI, your list includes dhammack's Github repository. He's also a contestant; that decision is up to Kaggle, but technically, you're both disqualified for sharing code outside of teams I guess. That's at their discretion though, of course.\n\n... Sorry, but if I'm going to lose, I'm only going to lose to people who play by the rules :(", "votes": -2 } ] }, { "id": 211444, "postDate": "2017-08-09T02:11:13.210Z", "content": "I want try it by tensorfow .
", "rawMarkdown": "I want try it by tensorfow .", "votes": -2 }, { "id": 255706, "postDate": "2017-12-09T21:35:20.800Z", "content": "I am using faster_rcnn_resnet101_coco pretrained models : http://download.tensorflow.org/models/object_detection/faster_rcnn_resnet101_coco_2017_11_08.tar.gz
", "rawMarkdown": "I am using faster_rcnn_resnet101_coco pretrained models : http://download.tensorflow.org/models/object_detection/faster_rcnn_resnet101_coco_2017_11_08.tar.gz" }, { "id": 254964, "postDate": "2017-12-07T23:51:19.350Z", "content": "Data sources already posted - Keras applications https://keras.io/applications/
", "rawMarkdown": "Data sources already posted - Keras applications https://keras.io/applications/" }, { "id": 253489, "postDate": "2017-12-05T02:56:56.117Z", "content": "FLIC Dataset: http://bensapp.github.io/datasets.html\nMPII Dataset: http://human-pose.mpi-inf.mpg.de/\nCOCO Dataset: http://cocodataset.org/
", "rawMarkdown": "FLIC Dataset: http://bensapp.github.io/datasets.html\nMPII Dataset: http://human-pose.mpi-inf.mpg.de/\nCOCO Dataset: http://cocodataset.org/" }, { "id": 253483, "postDate": "2017-12-05T02:24:43.637Z", "content": "May use InceptionV3, InceptionResNetV2, VGG16, VGG19, etc. (from https://keras.io/applications/). Also, pre-trained convolutional pose models from https://github.com/shihenw/convolutional-pose-machines-release. Link to original paper: https://arxiv.org/abs/1602.00134.
", "rawMarkdown": "May use InceptionV3, InceptionResNetV2, VGG16, VGG19, etc. (from https://keras.io/applications/). Also, pre-trained convolutional pose models from https://github.com/shihenw/convolutional-pose-machines-release. Link to original paper: https://arxiv.org/abs/1602.00134." }, { "id": 253405, "postDate": "2017-12-04T22:20:59.790Z", "content": "Pytorch pretrained models
", "rawMarkdown": "Pytorch pretrained models" }, { "id": 253396, "postDate": "2017-12-04T22:01:12.097Z", "content": "keras pretrained weights https://keras.io/applications/
", "rawMarkdown": "keras pretrained weights https://keras.io/applications/" }, { "id": 253382, "postDate": "2017-12-04T21:42:28.920Z", "content": "I may use the previously mentioned pre-trained weights for the various networks included in Keras (full list here: https://keras.io/applications/), e.g. those for InceptionV3, InceptionResNetV2, VGG16, VGG19, etc.
", "rawMarkdown": "I may use the previously mentioned pre-trained weights for the various networks included in Keras (full list here: https://keras.io/applications/), e.g. those for InceptionV3, InceptionResNetV2, VGG16, VGG19, etc." }, { "id": 253374, "postDate": "2017-12-04T21:17:11.337Z", "content": "Hi, we are using some of the trained weights from the models from here:\nhttps://github.com/soeaver/caffe-model\nhttps://github.com/shicai/DenseNet-Caffe
\n\nThank You\nZbigniew
", "rawMarkdown": "Hi, we are using some of the trained weights from the models from here:\nhttps://github.com/soeaver/caffe-model\nhttps://github.com/shicai/DenseNet-Caffe\n\nThank You\nZbigniew\n" }, { "id": 253279, "postDate": "2017-12-04T18:03:14.693Z", "content": "VGG16 from https://github.com/tflearn/tflearn/issues/267. And I am indebted to Mr. Farrar for his starter code and inspiration.
", "rawMarkdown": "VGG16 from https://github.com/tflearn/tflearn/issues/267. And I am indebted to Mr. Farrar for his starter code and inspiration." }, { "id": 253172, "postDate": "2017-12-04T14:59:44.167Z", "content": "I am using pre-trained object detection models found here: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md
\n\nKeras pre-trained models as mentioned by others: https://keras.io/applications/
\n\nI have built some code on top of this VGG16 model that I discovered while taking a Udacity class:\n'https://s3.amazonaws.com/content.udacity-data.com/nd101/vgg16.npy'
\n\nIs this publicly available data that can be used in this competition? If not, I will just switch to the pre-trained Keras models.
", "rawMarkdown": "I am using pre-trained object detection models found here: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md\n\nKeras pre-trained models as mentioned by others: https://keras.io/applications/\n\nI have built some code on top of this VGG16 model that I discovered while taking a Udacity class:\n'https://s3.amazonaws.com/content.udacity-data.com/nd101/vgg16.npy'\n\nIs this publicly available data that can be used in this competition? If not, I will just switch to the pre-trained Keras models." }, { "id": 252837, "postDate": "2017-12-04T01:44:36.660Z", "content": "I will be using pretrained models for use with Pytorch: http://pytorch.org/docs/0.2.0/torchvision/models.html https://github.com/Cadene/pretrained-models.pytorch
\n\nAnd also Keras pre-trained https://keras.io/applications/
\n\nMay use resnet / vgg weights trained on Imagenet from the Tensorflow Slim model zoo: https://github.com/tensorflow/models/tree/master/research/slim (Apache License 2.0)
\n\nMay also use a pretrained weights for DenseNet: https://github.com/titu1994/DenseNet https://github.com/tdeboissiere/DeepLearningImplementations
", "rawMarkdown": "I will be using pretrained models for use with Pytorch: http://pytorch.org/docs/0.2.0/torchvision/models.html https://github.com/Cadene/pretrained-models.pytorch\n\nAnd also Keras pre-trained https://keras.io/applications/\n\nMay use resnet / vgg weights trained on Imagenet from the Tensorflow Slim model zoo: https://github.com/tensorflow/models/tree/master/research/slim (Apache License 2.0)\n\nMay also use a pretrained weights for DenseNet: https://github.com/titu1994/DenseNet https://github.com/tdeboissiere/DeepLearningImplementations" }, { "id": 252828, "postDate": "2017-12-04T00:26:11.433Z", "content": "may use pre-trained models in Keras: https://keras.io/applications/
", "rawMarkdown": "may use pre-trained models in Keras: https://keras.io/applications/" }, { "id": 252735, "postDate": "2017-12-03T18:16:50.163Z", "content": "Using Keras pre-trained models: https://keras.io/applications/
", "rawMarkdown": "Using Keras pre-trained models: https://keras.io/applications/" }, { "id": 252716, "postDate": "2017-12-03T17:13:59.360Z", "content": "UFC101 Dataset: http://crcv.ucf.edu/data/UCF101.php
", "rawMarkdown": "UFC101 Dataset: http://crcv.ucf.edu/data/UCF101.php" }, { "id": 252705, "postDate": "2017-12-03T16:19:30.337Z", "content": "BAIR/BVLC GoogleNet Model: https://github.com/BVLC/caffe/tree/master/models/bvlc_googlenet
", "rawMarkdown": "BAIR/BVLC GoogleNet Model: https://github.com/BVLC/caffe/tree/master/models/bvlc_googlenet" }, { "id": 252681, "postDate": "2017-12-03T14:53:59.887Z", "content": "keras pretrained: https://keras.io/applications/
", "rawMarkdown": "keras pretrained: https://keras.io/applications/" }, { "id": 252679, "postDate": "2017-12-03T14:46:42.320Z", "rawMarkdown": "" }, { "id": 252659, "postDate": "2017-12-03T13:02:20.220Z", "content": "This comment previously stated that my submission was only initialized with pre-trained weights from the Keras project, but did not make use of them as submitted; I'm no longer certain enough of that to claim it, but I don't know how to delete a comment on Kaggle........
", "rawMarkdown": "This comment previously stated that my submission was only initialized with pre-trained weights from the Keras project, but did not make use of them as submitted; I'm no longer certain enough of that to claim it, but I don't know how to delete a comment on Kaggle........" }, { "id": 252528, "postDate": "2017-12-03T06:21:55.627Z", "content": "Keras pre-trained models: https://keras.io/applications/
\n\nAs well as weights for C3D converted to Keras from here provided as part of this GIST
", "rawMarkdown": "Keras pre-trained models: https://keras.io/applications/\n\nAs well as weights for [C3D][1] converted to Keras from [here][2] provided as part of [this GIST][3]\n\n [1]: http://vlg.cs.dartmouth.edu/c3d/\n [2]: https://drive.google.com/open?id=0BzmDUR5_3US0V1hjV3VvREJ3NGs\n [3]: https://gist.github.com/albertomontesg/d8b21a179c1e6cca0480ebdf292c34d2" }, { "id": 252344, "postDate": "2017-12-02T19:20:31.297Z", "content": "We'll be using\nKeras pre-trained https://keras.io/applications/\nas previous mentioned by many others.
", "rawMarkdown": "We'll be using\nKeras pre-trained https://keras.io/applications/\nas previous mentioned by many others.", "replies": [ { "id": 252845, "postDate": "2017-12-04T02:19:14.490Z", "content": "May also use a pretrained weights for DenseNet:\nhttps://github.com/titu1994/DenseNet\nhttps://github.com/tdeboissiere/DeepLearningImplementations
", "rawMarkdown": "May also use a pretrained weights for DenseNet:\nhttps://github.com/titu1994/DenseNet\nhttps://github.com/tdeboissiere/DeepLearningImplementations" } ] }, { "id": 252320, "postDate": "2017-12-02T18:48:08.840Z", "content": "Using pre-trained ImageNet models found here: https://github.com/BVLC/caffe/wiki/Model-Zoo https://github.com/Eniac-Xie/faster-rcnn-resnet
", "rawMarkdown": "Using pre-trained ImageNet models found here: https://github.com/BVLC/caffe/wiki/Model-Zoo https://github.com/Eniac-Xie/faster-rcnn-resnet" }, { "id": 251829, "postDate": "2017-12-01T18:56:18.367Z", "content": "Keras pre-trained models: https://keras.io/applications/ \nhttps://github.com/eldar/pose-tensorflow/blob/master/LICENSE
", "rawMarkdown": "Keras pre-trained models: https://keras.io/applications/ \nhttps://github.com/eldar/pose-tensorflow/blob/master/LICENSE" }, { "id": 251627, "postDate": "2017-12-01T13:56:27.177Z", "content": "use keras pre-trained models: https://keras.io/applications/, and pre-trained models from: https://github.com/lef-fan/cnn_finetune
", "rawMarkdown": "use keras pre-trained models: https://keras.io/applications/, and pre-trained models from: https://github.com/lef-fan/cnn_finetune\n" }, { "id": 251392, "postDate": "2017-12-01T05:19:14.723Z", "content": "Pretrained models for use with Pytorch:\nhttp://pytorch.org/docs/0.2.0/torchvision/models.html\nhttps://github.com/Cadene/pretrained-models.pytorch
", "rawMarkdown": "Pretrained models for use with Pytorch:\nhttp://pytorch.org/docs/0.2.0/torchvision/models.html\nhttps://github.com/Cadene/pretrained-models.pytorch\n" }, { "id": 251192, "postDate": "2017-11-30T20:10:56.890Z", "content": "Will use pretrained models in fastai library
", "rawMarkdown": "Will use pretrained models in fastai library" }, { "id": 250565, "postDate": "2017-11-30T05:26:00.780Z", "content": "I am also use the pre-trained networks at https://keras.io/applications/
", "rawMarkdown": "I am also use the pre-trained networks at https://keras.io/applications/" }, { "id": 250114, "postDate": "2017-11-29T20:19:22.243Z", "content": "Keras pre-trained models: https://keras.io/applications/
", "rawMarkdown": "Keras pre-trained models: https://keras.io/applications/" }, { "id": 250021, "postDate": "2017-11-29T16:48:59.583Z", "content": "I am also experimenting with keras pretrained networks at https://keras.io/applications/
", "rawMarkdown": "I am also experimenting with keras pretrained networks at https://keras.io/applications/" }, { "id": 249608, "postDate": "2017-11-28T20:21:39.857Z", "content": "may use pre-trained models in Keras: https://keras.io/applications/
", "rawMarkdown": "may use pre-trained models in Keras: https://keras.io/applications/" }, { "id": 249216, "postDate": "2017-11-28T01:06:08.567Z", "content": "Will use Resnet pre-trained models
", "rawMarkdown": "Will use Resnet pre-trained models" }, { "id": 248588, "postDate": "2017-11-26T15:00:46.173Z", "content": "Using various pretrained tf.slim models (e.g. InceptionVx and MobilenetV1x) from here and here
", "rawMarkdown": "Using various pretrained tf.slim models (e.g. InceptionVx and MobilenetV1x) from [here](https://github.com/tensorflow/models/tree/master/research/slim) and [here](https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet_v1.md)" }, { "id": 246804, "postDate": "2017-11-21T19:46:20.370Z", "content": "Keras pre-trained https://keras.io/applications/
", "rawMarkdown": "Keras pre-trained https://keras.io/applications/ " }, { "id": 245633, "postDate": "2017-11-19T04:28:44.103Z", "content": "Team Maven Wave may use inception weights found at:\ngs://cloud-ml-data/img/flower_photos/inception_v3_2016_08_28.ckpt
", "rawMarkdown": "Team Maven Wave may use inception weights found at:\ngs://cloud-ml-data/img/flower_photos/inception_v3_2016_08_28.ckpt" }, { "id": 242934, "postDate": "2017-11-13T01:46:40.033Z", "content": "I use Keras pre-trained model with ImageNet weights: https://keras.io/applications/ Do I have to specify what exact model I use? Do I have to provide the link to the weights? Both answers can help other competitors. So, is it required to provide them here?
", "rawMarkdown": "I use Keras pre-trained model with ImageNet weights: https://keras.io/applications/ Do I have to specify what exact model I use? Do I have to provide the link to the weights? Both answers can help other competitors. So, is it required to provide them here?", "replies": [ { "id": 252048, "postDate": "2017-12-02T06:02:18.513Z", "content": "Inception-ResNet V2 model for Keras: https://raw.githubusercontent.com/fchollet/keras/master/keras/applications/inception_resnet_v2.py\nImageNet pre-trained weights: https://github.com/fchollet/deep-learning-models/releases/download/v0.7/inception_resnet_v2_weights_tf_dim_ordering_tf_kernels.h5
", "rawMarkdown": "Inception-ResNet V2 model for Keras: https://raw.githubusercontent.com/fchollet/keras/master/keras/applications/inception_resnet_v2.py\nImageNet pre-trained weights: https://github.com/fchollet/deep-learning-models/releases/download/v0.7/inception_resnet_v2_weights_tf_dim_ordering_tf_kernels.h5" } ] }, { "id": 242914, "postDate": "2017-11-13T00:51:53.330Z", "content": "December 4, 2017 - Pre-trained models and external data posting deadline.
\n\nDecember 10, 2017 - Model upload & first stage deadline. This is the last day you may upload your model to be eligible for a prize.
\n\nI don't see any tools that allow to upload anything!
", "rawMarkdown": "**December 4, 2017 - Pre-trained models and external data posting deadline.**\n\n1. Where should we post pre-trained models?\n2. Where should we most manual labels? As I know, they are not considered as external data. Is it right?\n3. Should we post weights that are available publicly?\n\n**December 10, 2017 - Model upload & first stage deadline. This is the last day you may upload your model to be eligible for a prize.**\n\n4. Where should we upload models?\n5. How do you check minor differences (that are allowed, like path changes) in the solutions for stage 1 and stage 2? How can we be sure that you won't allow major differences to be present in the solutions for stage 2 for other participants?\n6. Do you retrain models when you check the solution? But any retrained model won't give the same results. So, do you accept pre-trained models as the final solution?\n\nI don't see any tools that allow to upload anything!", "replies": [ { "id": 242928, "postDate": "2017-11-13T01:25:54.223Z", "content": "Dmitry,
\n\nPre-trained models and external data can be posted in this thread, noted \"Use this thread to post sources of external data that you use as part of your approach.\" Any publicly available data used in crafting your model should be posted here.
\n\nModel upload will be made available in the coming weeks, and will be done through the Team tab. We do not open model upload until closer to the end of the competition to encourage version control :).
\n\nThanks for your patience!
", "rawMarkdown": "Dmitry,\n\nPre-trained models and external data can be posted in this thread, noted \"Use this thread to post sources of external data that you use as part of your approach.\" Any publicly available data used in crafting your model should be posted here.\n\nModel upload will be made available in the coming weeks, and will be done through the Team tab. We do not open model upload until closer to the end of the competition to encourage version control :).\n\nThanks for your patience!" }, { "id": 242929, "postDate": "2017-11-13T01:32:48.873Z", "content": "Thanks for your answer! But it's still unclear...
\n\nI hope, you don't want us to publish pre-trained models that generate final predictions here. Is it right that you are speaking only about other pre-trained models?
Is it right that we shouldn't publish extended labels on the training dataset here because they are not considered as external data? But where and when should we upload them then? (Probably, with the model on the Team tab?)
Re: \"publish extended labels\", I'm not entirely sure what's meant but wouldn't that be considered \"hand labeling\" which is expressly forbidden?
", "rawMarkdown": "Re: \"publish extended labels\", I'm not entirely sure what's meant but wouldn't that be considered \"hand labeling\" which is expressly forbidden?" }, { "id": 250356, "postDate": "2017-11-30T01:34:23.477Z", "content": "That means additional training data based on public training data. You can't do what you want with the training data. Hand labeling is apparently forbidden with the test data.
", "rawMarkdown": "That means additional training data based on public training data. You can't do what you want with the training data. Hand labeling is apparently forbidden with the test data." } ] }, { "id": 241075, "postDate": "2017-11-08T02:05:28.090Z", "content": "@wcukierski\nsorry for my double posting, but could you answer my question in this thread?\nhttps://www.kaggle.com/c/passenger-screening-algorithm-challenge/discussion/42791
", "rawMarkdown": "@wcukierski\nsorry for my double posting, but could you answer my question in this thread?\nhttps://www.kaggle.com/c/passenger-screening-algorithm-challenge/discussion/42791" }, { "id": 211416, "postDate": "2017-08-08T21:57:36.207Z", "content": "I plan to use Keras pretrained models with imagenet weights: https://keras.io/applications/
", "rawMarkdown": "I plan to use Keras pretrained models with imagenet weights: https://keras.io/applications/\n" }, { "id": 253351, "postDate": "2017-12-04T20:35:34.733Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 245833, "author_name": "Murray Miron", "author_url": "", "post_date": "2017-11-19T21:02:37.403000", "content": "If someone else has already posted external data that we use, is it necessary to disclose our own use of it as well? For example, I thought I'd try using some pretrained weights with ResNet to see if I get better performance than what I've put together: if it scores better, do I need to post my own disclosure of that, or is staying quiet because somebody else did fine?
\n\nThanks :)
", "votes": 1, "replies": [ { "id": 246177, "author_name": "Will Cukierski", "author_url": "", "post_date": "2017-11-20T16:37:25.337000", "content": "You don't have to repost if someone else has already posted it.
", "votes": 0, "replies": [] } ] }, { "id": 216491, "author_name": "Suchir Balaji", "author_url": "", "post_date": "2017-08-26T01:21:40.973000", "content": "Are we allowed to use external data in this competition? From reading the rules, it says
\n\n\n\n\nEXTERNAL DATA\n Unless otherwise expressly stated on the Competition Website, Participants must not use data other than the Data to develop and test their models and Submissions. Competition Sponsor reserves the right in its sole discretion to disqualify any Participant who Competition Sponsor discovers has undertaken or attempted to undertake the use of data other than the Data, or who uses the Data other than as permitted according to the Competition Website and in these Competition Rules, in the course of the Competition.
\n
but I don't see where it's \"expressly stated\" that we can use external data.
", "votes": 2, "replies": [ { "id": 216492, "author_name": "Suchir Balaji", "author_url": "", "post_date": "2017-08-26T01:33:10.160000", "content": "(oops, nevermind -- this is explicitly allowed earlier in the rules)
", "votes": 1, "replies": [] } ] }, { "id": 253430, "author_name": "Ivo Georgiev", "author_url": "", "post_date": "2017-12-04T23:53:05.240000", "content": "For the final model, I may use one or more of the following, data and/or pre-trained models:
\n\nPre-trained models and external datasets used zoo. The copyrights belong to the original authors. (I'm using only some of them of course (!) wherever allowed, but I thought the extended listing may be of use to others. Sorry for the length!):
\n\nExternal datasets (note: some also include published pre-trained benchmark models):
\n\nhttps://www.cs.toronto.edu/~kriz/cifar.html
\n\nhttps://deepmind.com/research/open-source/open-source-datasets/kinetics/
\n\nhttps://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI
\n\nhttp://cs.stanford.edu/people/karpathy/deepvideo/
\n\nhttp://human-pose.mpi-inf.mpg.de/
\n\nhttp://humanshape.mpi-inf.mpg.de/
\n\nhttp://homepages.inf.ed.ac.uk/rbf/CVonline/Imagedbase.htm
\n\nhttp://yann.lecun.com/exdb/mnist/
\n\n\n\nPre-trained models:
\n\nhttp://data.dmlc.ml/mxnet/models/
\n\nhttp://vlg.cs.dartmouth.edu/c3d/
\n\nhttp://humanshape.mpi-inf.mpg.de/#results
\n\nhttps://github.com/tensorflow/models/
\n\nhttps://mxnet.incubator.apache.org/model_zoo/
\n\nhttps://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/
\n\nhttps://github.com/dmlc/mxnet-model-gallery/
\n\nhttps://github.com/rai-project/mxnet/tree/master/builtin_models/
\n\nhttps://github.com/deepmind/kinetics-i3d/
\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/
\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/rcnn/
\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/image-classification/
\n\nhttps://github.com/tensorflow/models/tree/master/research/slim/
\n\nhttps://github.com/KeyKy/model-zoo/
\n\nhttps://github.com/xingyizhou/pose-hg-3d/
\n\nhttps://github.com/liuzhuang13/DenseNet/
\n\nhttps://github.com/SimJeg/FC-DenseNet/
\n\nhttps://github.com/ShichenLiu/CondenseNet/
\n\nhttps://github.com/msracver/Deformable-ConvNets/
\n\nhttps://github.com/szq0214/DSOD/
\n\nhttps://github.com/szagoruyko/functional-zoo/
\n\nhttps://github.com/JaggerYoung/C3D-mxnet/
\n\nhttps://github.com/soeaver/caffe-model/
\n\nhttps://github.com/rbgirshick/py-faster-rcnn/
\n\nhttps://github.com/YuwenXiong/py-R-FCN/
\n\nhttps://github.com/juliandewit/kaggle_ndsb2017/
\n\nhttps://github.com/dhammack/DSB2017/
\n\n\n\nhttps://github.com/bearpaw/PyraNet/
\n\nhttps://github.com/anewell/pose-hg-train/
", "votes": -1, "replies": [ { "id": 254329, "author_name": "Murray Miron", "author_url": "", "post_date": "2017-12-06T18:15:48.383000", "content": "> USP wrote\n> \n> > Pre-trained models and external datasets used zoo. The copyrights belong to the original authors. (I'm using only some of them of course (!) wherever allowed, but I thought the extended listing may be of use to others. Sorry for the length!):\n> \n> \n> ----------\n> \n> External datasets (note: some also include published pre-trained benchmark models):\n> \n> https://www.cs.toronto.edu/~kriz/cifar.html\n> \n> https://deepmind.com/research/open-source/open-source-datasets/kinetics/\n> \n> https://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI\n> \n> http://cs.stanford.edu/people/karpathy/deepvideo/\n> \n> http://human-pose.mpi-inf.mpg.de/\n> \n> http://humanshape.mpi-inf.mpg.de/\n> \n> http://homepages.inf.ed.ac.uk/rbf/CVonline/Imagedbase.htm\n> \n> http://yann.lecun.com/exdb/mnist/\n> \n> http://www.image-net.org\n> \n> ----------\n> \n> Pre-trained models:\n> \n> http://data.dmlc.ml/mxnet/models/\n> \n> http://vlg.cs.dartmouth.edu/c3d/\n> \n> http://humanshape.mpi-inf.mpg.de/#results\n> \n> https://github.com/tensorflow/models/\n> \n> https://mxnet.incubator.apache.org/model_zoo/\n> \n> https://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/\n> \n> https://github.com/dmlc/mxnet-model-gallery/\n> \n> https://github.com/rai-project/mxnet/tree/master/builtin_models/\n> \n> https://github.com/deepmind/kinetics-i3d/\n> \n> https://github.com/apache/incubator-mxnet/tree/master/example/\n> \n> https://github.com/apache/incubator-mxnet/tree/master/example/rcnn/\n> \n> https://github.com/apache/incubator-mxnet/tree/master/example/image-classification/\n> \n> https://github.com/tensorflow/models/tree/master/research/slim/\n> \n> https://github.com/KeyKy/model-zoo/\n> \n> https://github.com/xingyizhou/pose-hg-3d/\n> \n> https://github.com/liuzhuang13/DenseNet/\n> \n> https://github.com/SimJeg/FC-DenseNet/\n> \n> https://github.com/ShichenLiu/CondenseNet/\n> \n> https://github.com/msracver/Deformable-ConvNets/\n> \n> https://github.com/szq0214/DSOD/\n> \n> https://github.com/szagoruyko/functional-zoo/\n> \n> https://github.com/JaggerYoung/C3D-mxnet/\n> \n> https://github.com/soeaver/caffe-model/\n> \n> https://github.com/rbgirshick/py-faster-rcnn/\n> \n> https://github.com/YuwenXiong/py-R-FCN/\n> \n> https://github.com/juliandewit/kaggle_ndsb2017/\n> \n> https://github.com/dhammack/DSB2017/\n> \n> https://github.com/handong1587/handong1587.github.io/blob/master/_posts/deep_learning/2015-10-09-segmentation.md\n> \n> https://github.com/bearpaw/PyraNet/\n> \n> https://github.com/anewell/pose-hg-train/
\n\nFYI, your list includes dhammack's Github repository. He's also a contestant; that decision is up to Kaggle, but technically, you're both disqualified for sharing code outside of teams I guess. That's at their discretion though, of course.
\n\n... Sorry, but if I'm going to lose, I'm only going to lose to people who play by the rules :(
", "votes": -2, "replies": [] } ] }, { "id": 211444, "author_name": "yinguofeng", "author_url": "", "post_date": "2017-08-09T02:11:13.210000", "content": "I want try it by tensorfow .
", "votes": -2, "replies": [] }, { "id": 255706, "author_name": "Shiv Gowda", "author_url": "", "post_date": "2017-12-09T21:35:20.800000", "content": "I am using faster_rcnn_resnet101_coco pretrained models : http://download.tensorflow.org/models/object_detection/faster_rcnn_resnet101_coco_2017_11_08.tar.gz
", "votes": 0, "replies": [] }, { "id": 254964, "author_name": "James Thornton", "author_url": "", "post_date": "2017-12-07T23:51:19.350000", "content": "Data sources already posted - Keras applications https://keras.io/applications/
", "votes": 0, "replies": [] }, { "id": 253489, "author_name": "Jadiel", "author_url": "", "post_date": "2017-12-05T02:56:56.117000", "content": "FLIC Dataset: http://bensapp.github.io/datasets.html\nMPII Dataset: http://human-pose.mpi-inf.mpg.de/\nCOCO Dataset: http://cocodataset.org/
", "votes": 0, "replies": [] }, { "id": 253483, "author_name": "Brayan Jaramillo", "author_url": "", "post_date": "2017-12-05T02:24:43.637000", "content": "May use InceptionV3, InceptionResNetV2, VGG16, VGG19, etc. (from https://keras.io/applications/). Also, pre-trained convolutional pose models from https://github.com/shihenw/convolutional-pose-machines-release. Link to original paper: https://arxiv.org/abs/1602.00134.
", "votes": 0, "replies": [] }, { "id": 253405, "author_name": "Moejoe", "author_url": "", "post_date": "2017-12-04T22:20:59.790000", "content": "Pytorch pretrained models
", "votes": 0, "replies": [] }, { "id": 253396, "author_name": "Naoto Usuyama", "author_url": "", "post_date": "2017-12-04T22:01:12.097000", "content": "keras pretrained weights https://keras.io/applications/
", "votes": 0, "replies": [] }, { "id": 253382, "author_name": "havoc-io", "author_url": "", "post_date": "2017-12-04T21:42:28.920000", "content": "I may use the previously mentioned pre-trained weights for the various networks included in Keras (full list here: https://keras.io/applications/), e.g. those for InceptionV3, InceptionResNetV2, VGG16, VGG19, etc.
", "votes": 0, "replies": [] }, { "id": 253374, "author_name": "TensorFlight", "author_url": "", "post_date": "2017-12-04T21:17:11.337000", "content": "Hi, we are using some of the trained weights from the models from here:\nhttps://github.com/soeaver/caffe-model\nhttps://github.com/shicai/DenseNet-Caffe
\n\nThank You\nZbigniew
", "votes": 0, "replies": [] }, { "id": 253279, "author_name": "numericLee", "author_url": "", "post_date": "2017-12-04T18:03:14.693000", "content": "VGG16 from https://github.com/tflearn/tflearn/issues/267. And I am indebted to Mr. Farrar for his starter code and inspiration.
", "votes": 0, "replies": [] }, { "id": 253172, "author_name": "DeSpurious", "author_url": "", "post_date": "2017-12-04T14:59:44.167000", "content": "I am using pre-trained object detection models found here: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md
\n\nKeras pre-trained models as mentioned by others: https://keras.io/applications/
\n\nI have built some code on top of this VGG16 model that I discovered while taking a Udacity class:\n'https://s3.amazonaws.com/content.udacity-data.com/nd101/vgg16.npy'
\n\nIs this publicly available data that can be used in this competition? If not, I will just switch to the pre-trained Keras models.
", "votes": 0, "replies": [] }, { "id": 252837, "author_name": "YaGana Sheriff-Hussaini", "author_url": "", "post_date": "2017-12-04T01:44:36.660000", "content": "I will be using pretrained models for use with Pytorch: http://pytorch.org/docs/0.2.0/torchvision/models.html https://github.com/Cadene/pretrained-models.pytorch
\n\nAnd also Keras pre-trained https://keras.io/applications/
\n\nMay use resnet / vgg weights trained on Imagenet from the Tensorflow Slim model zoo: https://github.com/tensorflow/models/tree/master/research/slim (Apache License 2.0)
\n\nMay also use a pretrained weights for DenseNet: https://github.com/titu1994/DenseNet https://github.com/tdeboissiere/DeepLearningImplementations
", "votes": 0, "replies": [] }, { "id": 252828, "author_name": "Andrew", "author_url": "", "post_date": "2017-12-04T00:26:11.433000", "content": "may use pre-trained models in Keras: https://keras.io/applications/
", "votes": 0, "replies": [] }, { "id": 252735, "author_name": "emergent complexity", "author_url": "", "post_date": "2017-12-03T18:16:50.163000", "content": "Using Keras pre-trained models: https://keras.io/applications/
", "votes": 0, "replies": [] }, { "id": 252716, "author_name": "Christopher Goldsworthy", "author_url": "", "post_date": "2017-12-03T17:13:59.360000", "content": "UFC101 Dataset: http://crcv.ucf.edu/data/UCF101.php
", "votes": 0, "replies": [] }, { "id": 252705, "author_name": "serg14", "author_url": "", "post_date": "2017-12-03T16:19:30.337000", "content": "BAIR/BVLC GoogleNet Model: https://github.com/BVLC/caffe/tree/master/models/bvlc_googlenet
", "votes": 0, "replies": [] }, { "id": 252681, "author_name": "LivingProgram", "author_url": "", "post_date": "2017-12-03T14:53:59.887000", "content": "keras pretrained: https://keras.io/applications/
", "votes": 0, "replies": [] }, { "id": 252679, "author_name": "I can't feel my face.", "author_url": "", "post_date": "2017-12-03T14:46:42.320000", "content": "", "votes": 0, "replies": [] }, { "id": 252659, "author_name": "Murray Miron", "author_url": "", "post_date": "2017-12-03T13:02:20.220000", "content": "This comment previously stated that my submission was only initialized with pre-trained weights from the Keras project, but did not make use of them as submitted; I'm no longer certain enough of that to claim it, but I don't know how to delete a comment on Kaggle........
", "votes": 0, "replies": [] }, { "id": 252528, "author_name": "Earthos", "author_url": "", "post_date": "2017-12-03T06:21:55.627000", "content": "Keras pre-trained models: https://keras.io/applications/
\n\nAs well as weights for C3D converted to Keras from here provided as part of this GIST
", "votes": 0, "replies": [] }, { "id": 252344, "author_name": "Troy Retter", "author_url": "", "post_date": "2017-12-02T19:20:31.297000", "content": "We'll be using\nKeras pre-trained https://keras.io/applications/\nas previous mentioned by many others.
", "votes": 0, "replies": [ { "id": 252845, "author_name": "Troy Retter", "author_url": "", "post_date": "2017-12-04T02:19:14.490000", "content": "May also use a pretrained weights for DenseNet:\nhttps://github.com/titu1994/DenseNet\nhttps://github.com/tdeboissiere/DeepLearningImplementations
", "votes": 0, "replies": [] } ] }, { "id": 252320, "author_name": "Jocelyn Adams", "author_url": "", "post_date": "2017-12-02T18:48:08.840000", "content": "Using pre-trained ImageNet models found here: https://github.com/BVLC/caffe/wiki/Model-Zoo https://github.com/Eniac-Xie/faster-rcnn-resnet
", "votes": 0, "replies": [] }, { "id": 251829, "author_name": "Alon Daks", "author_url": "", "post_date": "2017-12-01T18:56:18.367000", "content": "Keras pre-trained models: https://keras.io/applications/ \nhttps://github.com/eldar/pose-tensorflow/blob/master/LICENSE
", "votes": 0, "replies": [] }, { "id": 251627, "author_name": "ploider", "author_url": "", "post_date": "2017-12-01T13:56:27.177000", "content": "use keras pre-trained models: https://keras.io/applications/, and pre-trained models from: https://github.com/lef-fan/cnn_finetune
", "votes": 0, "replies": [] }, { "id": 251392, "author_name": "", "author_url": "", "post_date": "2017-12-01T05:19:14.723000", "content": "", "votes": 0, "replies": [] }, { "id": 251192, "author_name": "", "author_url": "", "post_date": "2017-11-30T20:10:56.890000", "content": "", "votes": 0, "replies": [] }, { "id": 250565, "author_name": "", "author_url": "", "post_date": "2017-11-30T05:26:00.780000", "content": "", "votes": 0, "replies": [] }, { "id": 250114, "author_name": "", "author_url": "", "post_date": "2017-11-29T20:19:22.243000", "content": "", "votes": 0, "replies": [] }, { "id": 250021, "author_name": "", "author_url": "", "post_date": "2017-11-29T16:48:59.583000", "content": "", "votes": 0, "replies": [] }, { "id": 249608, "author_name": "", "author_url": "", "post_date": "2017-11-28T20:21:39.857000", "content": "", "votes": 0, "replies": [] }, { "id": 249216, "author_name": "", "author_url": "", "post_date": "2017-11-28T01:06:08.567000", "content": "", "votes": 0, "replies": [] }, { "id": 248588, "author_name": "", "author_url": "", "post_date": "2017-11-26T15:00:46.173000", "content": "", "votes": 0, "replies": [] }, { "id": 246804, "author_name": "", "author_url": "", "post_date": "2017-11-21T19:46:20.370000", "content": "", "votes": 0, "replies": [] }, { "id": 245633, "author_name": "", "author_url": "", "post_date": "2017-11-19T04:28:44.103000", "content": "", "votes": 0, "replies": [] }, { "id": 242934, "author_name": "", "author_url": "", "post_date": "2017-11-13T01:46:40.033000", "content": "", "votes": 0, "replies": [ { "id": 252048, "author_name": "", "author_url": "", "post_date": "2017-12-02T06:02:18.513000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 242914, "author_name": "", "author_url": "", "post_date": "2017-11-13T00:51:53.330000", "content": "", "votes": 0, "replies": [ { "id": 242928, "author_name": "", "author_url": "", "post_date": "2017-11-13T01:25:54.223000", "content": "", "votes": 0, "replies": [] }, { "id": 242929, "author_name": "", "author_url": "", "post_date": "2017-11-13T01:32:48.873000", "content": "", "votes": 0, "replies": [] }, { "id": 250350, "author_name": "", "author_url": "", "post_date": "2017-11-30T01:28:52.030000", "content": "", "votes": 0, "replies": [] }, { "id": 250356, "author_name": "", "author_url": "", "post_date": "2017-11-30T01:34:23.477000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 241075, "author_name": "", "author_url": "", "post_date": "2017-11-08T02:05:28.090000", "content": "", "votes": 0, "replies": [] }, { "id": 211416, "author_name": "", "author_url": "", "post_date": "2017-08-08T21:57:36.207000", "content": "", "votes": 0, "replies": [] }, { "id": 253351, "author_name": "", "author_url": "", "post_date": "2017-12-04T20:35:34.733000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "210932": "Use this thread to post sources of external data that you use as part of your approach.", "245833": "If someone else has already posted external data that we use, is it necessary to disclose our own use of it as well? For example, I thought I'd try using some pretrained weights with ResNet to see if I get better performance than what I've put together: if it scores better, do I need to post my own disclosure of that, or is staying quiet because somebody else did fine?\n\nThanks :)", "216491": "Are we allowed to use external data in this competition? From reading the rules, it says\n\n> EXTERNAL DATA\nUnless otherwise expressly stated on the Competition Website, Participants must not use data other than the Data to develop and test their models and Submissions. Competition Sponsor reserves the right in its sole discretion to disqualify any Participant who Competition Sponsor discovers has undertaken or attempted to undertake the use of data other than the Data, or who uses the Data other than as permitted according to the Competition Website and in these Competition Rules, in the course of the Competition.\n\nbut I don't see where it's \"expressly stated\" that we can use external data.", "253430": "For the final model, I may use one or more of the following, data and/or pre-trained models:\n\n 1. Keras (https://github.com/fchollet/keras/tree/master/keras/applications, https://keras.io/applications/, https://github.com/flyyufelix/cnn_finetune)\n 2. Tensorflow (https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md, https://github.com/tensorflow/models/tree/master/research/slim, https://github.com/tensorflow/models)\n 3. Stanford vision & geometry (https://github.com/charlesq34/3dcnn.torch, http://cvgl.stanford.edu/projects/pascal3d.html, http://cs229.stanford.edu/proj2012/AlAminChenLiuYeh-Learning3DHumanModels.pdf, http://cvgl.stanford.edu/resources.html)\n 4. Caffe/Caffe2 (https://caffe2.ai/docs/zoo.html)\n 5. C3D (http://vlg.cs.dartmouth.edu/c3d/, https://github.com/facebook/C3D)\n 6. mxnet (https://mxnet.incubator.apache.org/how_to/finetune.html)\n 7. FasterRCNN (https://github.com/rbgirshick/py-faster-rcnn)\n 8. Modelnet (http://modelnet.cs.princeton.edu/)\n 9. MVCNN (http://vis-www.cs.umass.edu/mvcnn/)\n 10. Marrnet (https://github.com/jiajunwu/marrnet, http://marrnet.csail.mit.edu/)\n 11. Shapenet (https://www.shapenet.org/)\n 12. Anthropometric db-s (http://store.sae.org/caesar/, http://openlab.psu.edu/design-tools-anthropometric-databases/)\n\n", "253414": "Pre-trained models and external datasets used zoo. The copyrights belong to the original authors. (I'm using only some of them of course (!) wherever allowed, but I thought the extended listing may be of use to others. Sorry for the length!):\n\n\n----------\n\nExternal datasets (note: some also include published pre-trained benchmark models):\n\nhttps://www.cs.toronto.edu/~kriz/cifar.html\n\nhttps://deepmind.com/research/open-source/open-source-datasets/kinetics/\n\nhttps://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI\n\nhttp://cs.stanford.edu/people/karpathy/deepvideo/\n\nhttp://human-pose.mpi-inf.mpg.de/\n\nhttp://humanshape.mpi-inf.mpg.de/\n\nhttp://homepages.inf.ed.ac.uk/rbf/CVonline/Imagedbase.htm\n\nhttp://yann.lecun.com/exdb/mnist/\n\nhttp://www.image-net.org\n\n----------\n\nPre-trained models:\n\nhttp://data.dmlc.ml/mxnet/models/\n\nhttp://vlg.cs.dartmouth.edu/c3d/\n\nhttp://humanshape.mpi-inf.mpg.de/#results\n\nhttps://github.com/tensorflow/models/\n\nhttps://mxnet.incubator.apache.org/model_zoo/\n\nhttps://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/\n\nhttps://github.com/dmlc/mxnet-model-gallery/\n\nhttps://github.com/rai-project/mxnet/tree/master/builtin_models/\n\nhttps://github.com/deepmind/kinetics-i3d/\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/rcnn/\n\nhttps://github.com/apache/incubator-mxnet/tree/master/example/image-classification/\n\nhttps://github.com/tensorflow/models/tree/master/research/slim/\n\nhttps://github.com/KeyKy/model-zoo/\n\nhttps://github.com/xingyizhou/pose-hg-3d/\n\nhttps://github.com/liuzhuang13/DenseNet/\n\nhttps://github.com/SimJeg/FC-DenseNet/\n\nhttps://github.com/ShichenLiu/CondenseNet/\n\nhttps://github.com/msracver/Deformable-ConvNets/\n\nhttps://github.com/szq0214/DSOD/\n\nhttps://github.com/szagoruyko/functional-zoo/\n\nhttps://github.com/JaggerYoung/C3D-mxnet/\n\nhttps://github.com/soeaver/caffe-model/\n\nhttps://github.com/rbgirshick/py-faster-rcnn/\n\nhttps://github.com/YuwenXiong/py-R-FCN/\n\nhttps://github.com/juliandewit/kaggle_ndsb2017/\n\nhttps://github.com/dhammack/DSB2017/\n\nhttps://github.com/handong1587/handong1587.github.io/blob/master/_posts/deep_learning/2015-10-09-segmentation.md\n\nhttps://github.com/bearpaw/PyraNet/\n\nhttps://github.com/anewell/pose-hg-train/", "211444": "I want try it by tensorfow .", "255706": "I am using faster_rcnn_resnet101_coco pretrained models : http://download.tensorflow.org/models/object_detection/faster_rcnn_resnet101_coco_2017_11_08.tar.gz", "254964": "Data sources already posted - Keras applications https://keras.io/applications/", "253489": "FLIC Dataset: http://bensapp.github.io/datasets.html\nMPII Dataset: http://human-pose.mpi-inf.mpg.de/\nCOCO Dataset: http://cocodataset.org/", "253483": "May use InceptionV3, InceptionResNetV2, VGG16, VGG19, etc. (from https://keras.io/applications/). Also, pre-trained convolutional pose models from https://github.com/shihenw/convolutional-pose-machines-release. Link to original paper: https://arxiv.org/abs/1602.00134.", "253405": "Pytorch pretrained models", "253396": "keras pretrained weights https://keras.io/applications/", "253382": "I may use the previously mentioned pre-trained weights for the various networks included in Keras (full list here: https://keras.io/applications/), e.g. those for InceptionV3, InceptionResNetV2, VGG16, VGG19, etc.", "253374": "Hi, we are using some of the trained weights from the models from here:\nhttps://github.com/soeaver/caffe-model\nhttps://github.com/shicai/DenseNet-Caffe\n\nThank You\nZbigniew\n", "253279": "VGG16 from https://github.com/tflearn/tflearn/issues/267. And I am indebted to Mr. Farrar for his starter code and inspiration.", "253172": "I am using pre-trained object detection models found here: https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md\n\nKeras pre-trained models as mentioned by others: https://keras.io/applications/\n\nI have built some code on top of this VGG16 model that I discovered while taking a Udacity class:\n'https://s3.amazonaws.com/content.udacity-data.com/nd101/vgg16.npy'\n\nIs this publicly available data that can be used in this competition? If not, I will just switch to the pre-trained Keras models.", "252837": "I will be using pretrained models for use with Pytorch: http://pytorch.org/docs/0.2.0/torchvision/models.html https://github.com/Cadene/pretrained-models.pytorch\n\nAnd also Keras pre-trained https://keras.io/applications/\n\nMay use resnet / vgg weights trained on Imagenet from the Tensorflow Slim model zoo: https://github.com/tensorflow/models/tree/master/research/slim (Apache License 2.0)\n\nMay also use a pretrained weights for DenseNet: https://github.com/titu1994/DenseNet https://github.com/tdeboissiere/DeepLearningImplementations", "252828": "may use pre-trained models in Keras: https://keras.io/applications/", "252735": "Using Keras pre-trained models: https://keras.io/applications/", "252716": "UFC101 Dataset: http://crcv.ucf.edu/data/UCF101.php", "252705": "BAIR/BVLC GoogleNet Model: https://github.com/BVLC/caffe/tree/master/models/bvlc_googlenet", "252681": "keras pretrained: https://keras.io/applications/", "252679": "", "252659": "This comment previously stated that my submission was only initialized with pre-trained weights from the Keras project, but did not make use of them as submitted; I'm no longer certain enough of that to claim it, but I don't know how to delete a comment on Kaggle........", "252528": "Keras pre-trained models: https://keras.io/applications/\n\nAs well as weights for [C3D][1] converted to Keras from [here][2] provided as part of [this GIST][3]\n\n [1]: http://vlg.cs.dartmouth.edu/c3d/\n [2]: https://drive.google.com/open?id=0BzmDUR5_3US0V1hjV3VvREJ3NGs\n [3]: https://gist.github.com/albertomontesg/d8b21a179c1e6cca0480ebdf292c34d2", "252344": "We'll be using\nKeras pre-trained https://keras.io/applications/\nas previous mentioned by many others.", "252320": "Using pre-trained ImageNet models found here: https://github.com/BVLC/caffe/wiki/Model-Zoo https://github.com/Eniac-Xie/faster-rcnn-resnet", "251829": "1. VGG16 Weights found here: https://www.cs.toronto.edu/~frossard/post/vgg16/. These weights are converted from the caffe weights here: https://gist.github.com/ksimonyan/211839e770f7b538e2d8 (The license references this link: http://www.robots.ox.ac.uk/~vgg/research/very_deep/ which is Creative Commons Attribution 4.0). The conversion script is found here: https://github.com/ethereon/caffe-tensorflow (MIT License). These weights are trained on Imagenet.\n2. May use resnet / vgg weights trained on Imagenet from the Tensorflow Slim model zoo: https://github.com/tensorflow/models/tree/master/research/slim (Apache License 2.0)", "251812": "Keras pre-trained models: https://keras.io/applications/ \nhttps://github.com/eldar/pose-tensorflow/blob/master/LICENSE", "251627": "use keras pre-trained models: https://keras.io/applications/, and pre-trained models from: https://github.com/lef-fan/cnn_finetune\n", "251392": "Pretrained models for use with Pytorch:\nhttp://pytorch.org/docs/0.2.0/torchvision/models.html\nhttps://github.com/Cadene/pretrained-models.pytorch\n", "251192": "Will use pretrained models in fastai library", "250565": "I am also use the pre-trained networks at https://keras.io/applications/", "250114": "Keras pre-trained models: https://keras.io/applications/", "250021": "I am also experimenting with keras pretrained networks at https://keras.io/applications/", "249608": "may use pre-trained models in Keras: https://keras.io/applications/", "249216": "Will use Resnet pre-trained models", "248588": "Using various pretrained tf.slim models (e.g. InceptionVx and MobilenetV1x) from [here](https://github.com/tensorflow/models/tree/master/research/slim) and [here](https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet_v1.md)", "246804": "Keras pre-trained https://keras.io/applications/ ", "245633": "Team Maven Wave may use inception weights found at:\ngs://cloud-ml-data/img/flower_photos/inception_v3_2016_08_28.ckpt", "242934": "I use Keras pre-trained model with ImageNet weights: https://keras.io/applications/ Do I have to specify what exact model I use? Do I have to provide the link to the weights? Both answers can help other competitors. So, is it required to provide them here?", "242914": "**December 4, 2017 - Pre-trained models and external data posting deadline.**\n\n1. Where should we post pre-trained models?\n2. Where should we most manual labels? As I know, they are not considered as external data. Is it right?\n3. Should we post weights that are available publicly?\n\n**December 10, 2017 - Model upload & first stage deadline. This is the last day you may upload your model to be eligible for a prize.**\n\n4. Where should we upload models?\n5. How do you check minor differences (that are allowed, like path changes) in the solutions for stage 1 and stage 2? How can we be sure that you won't allow major differences to be present in the solutions for stage 2 for other participants?\n6. Do you retrain models when you check the solution? But any retrained model won't give the same results. So, do you accept pre-trained models as the final solution?\n\nI don't see any tools that allow to upload anything!", "241075": "@wcukierski\nsorry for my double posting, but could you answer my question in this thread?\nhttps://www.kaggle.com/c/passenger-screening-algorithm-challenge/discussion/42791", "211416": "I plan to use Keras pretrained models with imagenet weights: https://keras.io/applications/\n", "253351": "" } }