You should feel free to use pretrained models for your entries in this competition, but you must post them here before the competition's entry deadline.
", "messages": [ { "id": 171108, "postDate": "2017-03-28T17:32:15.603Z", "content": "Is it for a model pretrained by me or using pretrained model by anyone, e.g. VGG?\nDo you have any requirement for the model: where to host the file, file format etc
", "rawMarkdown": "Is it for a model pretrained by me or using pretrained model by anyone, e.g. VGG?\nDo you have any requirement for the model: where to host the file, file format etc", "votes": 3, "replies": [ { "id": 175750, "postDate": "2017-04-17T10:57:02.027Z", "content": "I would suggest you to read the \"Pretrained models thread\" for other recent Kaggle competitions. It will solve the most of your doubts about how to proceed. You've got one here;\nhttps://www.kaggle.com/c/the-nature-conservancy-fisheries-monitoring/discussion/25428
", "rawMarkdown": "I would suggest you to read the \"Pretrained models thread\" for other recent Kaggle competitions. It will solve the most of your doubts about how to proceed. You've got one here;\nhttps://www.kaggle.com/c/the-nature-conservancy-fisheries-monitoring/discussion/25428" }, { "id": 192695, "postDate": "2017-06-14T13:42:10.360Z", "content": "Hi, Igor, did you get an answer?
", "rawMarkdown": "Hi, Igor, did you get an answer?" } ] }, { "id": 170868, "postDate": "2017-03-27T21:14:41.337Z", "content": "You should feel free to use pretrained models for your entries in this competition, but you must post them here before the competition's entry deadline.
", "rawMarkdown": "You should feel free to use pretrained models for your entries in this competition, but you must post them here before the competition's entry deadline.", "votes": 4 }, { "id": 192791, "postDate": "2017-06-14T20:06:44.290Z", "content": "torchvision models pretrained on ImageNet: http://pytorch.org/docs/torchvision/models.html
", "rawMarkdown": "torchvision models pretrained on ImageNet: http://pytorch.org/docs/torchvision/models.html", "votes": 1 }, { "id": 176391, "postDate": "2017-04-20T09:50:19.927Z", "content": "I guess that in this competition neither VGG nor IMAGENET will help much. They just might with a lot of finetuning. The best idea I found so far is using nearest neighbors first on train pictures numbers ( as lions tend to group in same numbers on islands). Puting those in folders, training conv network in recognising them and testing. Does it sound sane? I guess dot approach will be neither doable nor any help for real life solution for NOAA
", "rawMarkdown": "I guess that in this competition neither VGG nor IMAGENET will help much. They just might with a lot of finetuning. The best idea I found so far is using nearest neighbors first on train pictures numbers ( as lions tend to group in same numbers on islands). Puting those in folders, training conv network in recognising them and testing. Does it sound sane? I guess dot approach will be neither doable nor any help for real life solution for NOAA ", "votes": 1 }, { "id": 176278, "postDate": "2017-04-19T19:23:53.510Z", "content": "Since the thread here is not quite understood. As an example you may want to use one of these pretrained models and tweak it to your desires.
\n\nkeras inception v3 pretrained on imagenet
", "rawMarkdown": "keras inception v3 pretrained on imagenet" }, { "id": 196256, "postDate": "2017-06-26T21:06:08.910Z", "content": "keras pretrained resnet50 https://github.com/fchollet/deep-learning-models/releases/download/v0.2/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5
", "rawMarkdown": "keras pretrained resnet50 https://github.com/fchollet/deep-learning-models/releases/download/v0.2/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5" }, { "id": 195679, "postDate": "2017-06-24T14:02:01.620Z", "content": "", "rawMarkdown": "[darknet19_448.weights][1]\n\n [1]: https://pjreddie.com/darknet/" }, { "id": 195166, "postDate": "2017-06-22T21:50:51.487Z", "content": "Faster RCNN pre-trained on VOC 2007 challenge data set
", "rawMarkdown": "Faster RCNN pre-trained on VOC 2007 challenge data set" }, { "id": 195137, "postDate": "2017-06-22T20:12:23.490Z", "content": "", "rawMarkdown": "[TF slim mobilenet][1]\n\n\n [1]: https://github.com/tensorflow/models/tree/master/slim#Pretrained" }, { "id": 194891, "postDate": "2017-06-22T05:04:22.887Z", "content": "Keras applications https://keras.io/applications/
", "rawMarkdown": "Keras applications https://keras.io/applications/" }, { "id": 194872, "postDate": "2017-06-22T03:01:33.747Z", "content": "", "rawMarkdown": "[keras application][1]\n\n\n [1]: https://keras.io/applications/" }, { "id": 194814, "postDate": "2017-06-21T21:19:40.933Z", "content": "Keras VGG16 with weights pre-trained on ImageNet.
", "rawMarkdown": "[Keras VGG16](https://github.com/fchollet/keras/blob/master/keras/applications/vgg16.py) with weights pre-trained on ImageNet." }, { "id": 194531, "postDate": "2017-06-20T23:41:43.847Z", "content": "Initially I thought as long as someone posted the data source (imagenet, posted a couple of times) in here it's acceptable to use it without commenting here, but looking at the comments everyone seems to post what they use.
\n\nSo I think I just missed the deadline, but I was planning to use Inception v1 pretrained on ImageNet from here https://github.com/tensorflow/models/tree/master/slim (weights http://download.tensorflow.org/models/inception_v1_2016_08_28.tar.gz). Is this still acceptable? If not I just won't use it.
", "rawMarkdown": "Initially I thought as long as someone posted the data source (imagenet, posted a couple of times) in here it's acceptable to use it without commenting here, but looking at the comments everyone seems to post what they use.\n\nSo I think I just missed the deadline, but I was planning to use Inception v1 pretrained on ImageNet from here https://github.com/tensorflow/models/tree/master/slim (weights http://download.tensorflow.org/models/inception_v1_2016_08_28.tar.gz). Is this still acceptable? If not I just won't use it." }, { "id": 194512, "postDate": "2017-06-20T22:11:19.007Z", "content": "Keras VGG16 pretrained on imagenet
", "rawMarkdown": "Keras VGG16 pretrained on imagenet" }, { "id": 194508, "postDate": "2017-06-20T21:57:06.820Z", "content": "https://keras.io/applications/\nhttps://github.com/kentsommer/keras-inceptionV4\nhttps://gist.github.com/flyyufelix/7e2eafb149f72f4d38dd661882c554a6
", "rawMarkdown": "https://keras.io/applications/\nhttps://github.com/kentsommer/keras-inceptionV4\nhttps://gist.github.com/flyyufelix/7e2eafb149f72f4d38dd661882c554a6" }, { "id": 194448, "postDate": "2017-06-20T16:33:51.020Z", "content": "https://keras.io/applications/
\n\nhttps://github.com/rykov8/ssd_keras
", "rawMarkdown": "https://keras.io/applications/ \n\nhttps://github.com/rykov8/ssd_keras" }, { "id": 194438, "postDate": "2017-06-20T15:56:38.323Z", "content": "VGG and ResNet
", "rawMarkdown": "VGG and ResNet" }, { "id": 194436, "postDate": "2017-06-20T15:49:24.780Z", "content": "VGG16 Keras pretrained
", "rawMarkdown": "VGG16 Keras pretrained" }, { "id": 194412, "postDate": "2017-06-20T13:07:08.997Z", "content": "https://github.com/tensorflow/models/tree/master/slim#Pretrained
\n\nhttps://keras.io/applications/
", "rawMarkdown": "https://github.com/tensorflow/models/tree/master/slim#Pretrained\n\nhttps://keras.io/applications/" }, { "id": 194251, "postDate": "2017-06-19T21:24:09Z", "content": "", "rawMarkdown": "[Keras pretrained][1] \n\n\n [1]: https://keras.io/applications/#xception" }, { "id": 194130, "postDate": "2017-06-19T12:49:41.733Z", "content": "http://download.tensorflow.org/models/object_detection/ssd_inception_v2_coco_11_06_2017.tar.gz
\n\nhttp://download.tensorflow.org/models/object_detection/rfcn_resnet101_coco_11_06_2017.tar.gz
\n\nhttp://download.tensorflow.org/models/object_detection/faster_rcnn_resnet101_coco_11_06_2017.tar.gz
\n\n\n\nhttps://pjreddie.com/media/files/yolo-voc.weights
\n\nSSD500 ILSVRC\nhttps://drive.google.com/open?id=0BzKzrI_SkD1_X2ZCLVgwLTgzaTQ
\n\nSSD512 07++12++COCO\nhttps://drive.google.com/open?id=0BzKzrI_SkD1_NVVNdWdYNEh1WTA
\n\nhttp://cs.unc.edu/~wliu/projects/SSD/ZF_conv_reduced.caffemodel
", "rawMarkdown": "http://download.tensorflow.org/models/object_detection/ssd_inception_v2_coco_11_06_2017.tar.gz\n\nhttp://download.tensorflow.org/models/object_detection/rfcn_resnet101_coco_11_06_2017.tar.gz\n\nhttp://download.tensorflow.org/models/object_detection/faster_rcnn_resnet101_coco_11_06_2017.tar.gz\n\nhttp://download.tensorflow.org/models/object_detection/faster_rcnn_inception_resnet_v2_atrous_coco_11_06_2017.tar.gz\n\nhttps://pjreddie.com/media/files/yolo-voc.weights\n\nSSD500 ILSVRC\nhttps://drive.google.com/open?id=0BzKzrI_SkD1_X2ZCLVgwLTgzaTQ\n\nSSD512 07++12++COCO\nhttps://drive.google.com/open?id=0BzKzrI_SkD1_NVVNdWdYNEh1WTA\n\nhttp://cs.unc.edu/~wliu/projects/SSD/ZF_conv_reduced.caffemodel" }, { "id": 194085, "postDate": "2017-06-19T10:21:53.500Z", "content": "Hi, we are using pretrained VGG16
", "rawMarkdown": "Hi, we are using pretrained VGG16" }, { "id": 194016, "postDate": "2017-06-19T05:50:51.670Z", "content": "Keras VGG16 trained on Imagenet
", "rawMarkdown": "Keras VGG16 trained on Imagenet", "replies": [ { "id": 194018, "postDate": "2017-06-19T05:51:51.970Z", "content": "Fast RCNN based on AlexNet trained on Imagenet
", "rawMarkdown": "Fast RCNN based on AlexNet trained on Imagenet" } ] }, { "id": 193947, "postDate": "2017-06-18T22:14:13.040Z", "content": "Tensorflow slim pretrained models: https://github.com/tensorflow/models/tree/master/slim#Pretrained
", "rawMarkdown": "Tensorflow slim pretrained models: https://github.com/tensorflow/models/tree/master/slim#Pretrained" }, { "id": 193933, "postDate": "2017-06-18T21:14:48.107Z", "content": "Will try:
\n\nKeras pretrained models : https://keras.io/applications/
\n\nTensorflow pretrained models: https://github.com/tensorflow/models/tree/master/slim#Pretrained
", "rawMarkdown": "Will try:\n\nKeras pretrained models : https://keras.io/applications/\n\nTensorflow pretrained models: https://github.com/tensorflow/models/tree/master/slim#Pretrained\n" }, { "id": 193852, "postDate": "2017-06-18T12:49:27.607Z", "content": "weights pretrained on IMAGENET. here is the link: https://pjreddie.com/media/files/darknet19_448.conv.23
", "rawMarkdown": "weights pretrained on IMAGENET. here is the link: https://pjreddie.com/media/files/darknet19_448.conv.23" }, { "id": 193620, "postDate": "2017-06-17T06:11:51.483Z", "content": "Keras VGG16 trained on Imagenet (Keras pretrained)
", "rawMarkdown": "Keras VGG16 trained on Imagenet (Keras pretrained)", "replies": [ { "id": 194017, "postDate": "2017-06-19T05:51:25.277Z", "rawMarkdown": "" } ] }, { "id": 193224, "postDate": "2017-06-15T20:36:06.857Z", "content": "ResNet, Inception, Inception-ResNet
", "rawMarkdown": "ResNet, Inception, Inception-ResNet" }, { "id": 192891, "postDate": "2017-06-15T03:25:12.187Z", "content": "I'm using VGG-16 from https://keras.io/applications/
", "rawMarkdown": "I'm using VGG-16 from https://keras.io/applications/" }, { "id": 192767, "postDate": "2017-06-14T18:17:48.530Z", "content": "VGG16 from https://keras.io/applications/
", "rawMarkdown": "VGG16 from https://keras.io/applications/" }, { "id": 192739, "postDate": "2017-06-14T16:31:43.190Z", "content": "I am using the pretrained models with imagenet weights from Keras: https://keras.io/applications/
", "rawMarkdown": "I am using the pretrained models with imagenet weights from Keras: https://keras.io/applications/" }, { "id": 192726, "postDate": "2017-06-14T15:24:13.760Z", "content": "VGG16 and Resnet50
", "rawMarkdown": "VGG16 and Resnet50" }, { "id": 192724, "postDate": "2017-06-14T15:18:24.810Z", "content": "I'm using keras and ssd https://github.com/rykov8/ssd_keras with the weights pretrained on VOC2007 (https://mega.nz/#F!7RowVLCL!q3cEVRK9jyOSB9el3SssIA)
", "rawMarkdown": "I'm using keras and ssd https://github.com/rykov8/ssd_keras with the weights pretrained on VOC2007 (https://mega.nz/#F!7RowVLCL!q3cEVRK9jyOSB9el3SssIA)" }, { "id": 192474, "postDate": "2017-06-13T16:47:38.513Z", "content": "using keras pretrained models. \nAdditionally using mobilenet I pretrained on imagenet. https://github.com/danzelmo/mobilenet
\n\nEdit: also https://github.com/tensorflow/models/tree/master/slim#Pretrained, and \nhttps://github.com/tensorflow/models/blob/master/object_detection/g3doc/detection_model_zoo.md
", "rawMarkdown": "using keras pretrained models. \nAdditionally using mobilenet I pretrained on imagenet. https://github.com/danzelmo/mobilenet\n\nEdit: also https://github.com/tensorflow/models/tree/master/slim#Pretrained, and \nhttps://github.com/tensorflow/models/blob/master/object_detection/g3doc/detection_model_zoo.md" }, { "id": 192023, "postDate": "2017-06-12T13:38:37.250Z", "content": "Resnet50.https://github.com/tensorflow/models/tree/master/slim
", "rawMarkdown": "Resnet50.https://github.com/tensorflow/models/tree/master/slim" }, { "id": 191920, "postDate": "2017-06-12T05:26:24.070Z", "content": "Hi, we are using Inception V3
", "rawMarkdown": "Hi, we are using Inception V3" }, { "id": 191632, "postDate": "2017-06-11T06:23:41.510Z", "content": "I'm using keras applications https://keras.io/applications/\nand ssd keras port https://github.com/rykov8/ssd_keras
", "rawMarkdown": "I'm using keras applications https://keras.io/applications/\nand ssd keras port https://github.com/rykov8/ssd_keras" }, { "id": 191166, "postDate": "2017-06-09T14:00:03.070Z", "content": "Inception-v4 from https://github.com/tensorflow/models/tree/master/slim
", "rawMarkdown": "Inception-v4 from https://github.com/tensorflow/models/tree/master/slim" }, { "id": 190295, "postDate": "2017-06-07T13:00:36.687Z", "content": "I use torch ResNet.\nhttps://github.com/facebook/fb.resnet.torch/tree/master/pretrained
", "rawMarkdown": "I use torch ResNet.\nhttps://github.com/facebook/fb.resnet.torch/tree/master/pretrained\n" }, { "id": 189450, "postDate": "2017-06-06T04:06:43.743Z", "content": "Pretrained model: VGG16 trained on Imagenet.
", "rawMarkdown": "Pretrained model: VGG16 trained on Imagenet." }, { "id": 188351, "postDate": "2017-06-02T17:58:38.960Z", "content": "https://keras.io/applications/
", "rawMarkdown": "https://keras.io/applications/" }, { "id": 188350, "postDate": "2017-06-02T17:56:44.770Z", "content": "RESNET (-50, -101) pretrained on ImageNet
", "rawMarkdown": "RESNET (-50, -101) pretrained on ImageNet" }, { "id": 184899, "postDate": "2017-05-23T13:10:22.480Z", "content": "Pretrained ZF
", "rawMarkdown": "Pretrained ZF" }, { "id": 183276, "postDate": "2017-05-17T16:15:19.097Z", "content": "https://keras.io/applications/ also
", "rawMarkdown": "https://keras.io/applications/ also" }, { "id": 183265, "postDate": "2017-05-17T15:33:54.180Z", "content": "Keras applications https://keras.io/applications/
", "rawMarkdown": "Keras applications https://keras.io/applications/" }, { "id": 182789, "postDate": "2017-05-15T21:11:44.587Z", "content": "VGG16
", "rawMarkdown": "VGG16" }, { "id": 182576, "postDate": "2017-05-14T17:38:37.180Z", "rawMarkdown": "" }, { "id": 182481, "postDate": "2017-05-14T03:41:30.090Z", "content": "VGG16
", "rawMarkdown": "VGG16" }, { "id": 182011, "postDate": "2017-05-11T23:53:11.500Z", "content": "keras frcnn
\ndarknet(yolo) keras
\nssd keras
Same here.
", "rawMarkdown": "Same here." } ] }, { "id": 180539, "postDate": "2017-05-05T19:51:43.940Z", "content": "Using the weights in the respective files posted by Felix Yu here : https://github.com/flyyufelix/cnn_finetune/
", "rawMarkdown": "Using the weights in the respective files posted by Felix Yu here : https://github.com/flyyufelix/cnn_finetune/" }, { "id": 180506, "postDate": "2017-05-05T18:01:16.377Z", "content": "Pretrained VGG16
", "rawMarkdown": "Pretrained VGG16" }, { "id": 180345, "postDate": "2017-05-05T02:04:55.747Z", "content": "Keras VGG16
", "rawMarkdown": "Keras [VGG16][1]\n\n\n [1]: https://github.com/fchollet/keras/blob/master/keras/applications/vgg16.py" }, { "id": 180285, "postDate": "2017-05-04T18:45:42.720Z", "content": "I'm using keras applications\nhttps://keras.io/applications/
", "rawMarkdown": "I'm using keras applications\nhttps://keras.io/applications/", "replies": [ { "id": 180787, "postDate": "2017-05-07T07:30:50.270Z", "content": "I'm using the keras applications as well
", "rawMarkdown": "I'm using the keras applications as well" } ] }, { "id": 180069, "postDate": "2017-05-04T00:10:28.220Z", "content": "Pretrained VGG , RESNET and VGG_reduced models.
", "rawMarkdown": "Pretrained [VGG][1] , [RESNET][2] and [VGG_reduced][3] models. \n\n\n [1]: https://github.com/pytorch/vision/blob/master/torchvision/models/vgg.py\n [2]: https://github.com/pytorch/vision/blob/master/torchvision/models/resnet.py\n [3]: https://github.com/zhreshold/mxnet-ssd/releases/download/v0.2-beta/vgg16_reduced.zip" }, { "id": 171051, "postDate": "2017-03-28T13:39:52.107Z", "content": "Clarification: Do you want only the pretrained model used, or the pretrained + any backends added
", "rawMarkdown": "Clarification: Do you want only the pretrained model used, or the pretrained + any backends added\n\n" }, { "id": 193568, "postDate": "2017-06-16T23:24:31.277Z", "rawMarkdown": "", "isDeleted": true }, { "id": 179968, "postDate": "2017-05-03T13:59:43.097Z", "rawMarkdown": "", "isDeleted": true, "replies": [ { "id": 180011, "postDate": "2017-05-03T17:35:04.880Z", "content": "I want to use this opportunity to ask what about rotating, scaling, contrast increasing/decreasing and any other (non-)affine transformations applied to images from miss matched list included in dataset. Is it able or unable to do this by myself?
\n\nAs mentioned in Rules
\n\n\n", "rawMarkdown": "I want to use this opportunity to ask what about rotating, scaling, contrast increasing/decreasing and any other (non-)affine transformations applied to images from miss matched list included in dataset. Is it able or unable to do this by myself?\n\n\nAs mentioned in __Rules__\n> Unless otherwise permitted by the terms of the Competition Website, Participants must use the Data solely for the purpose and duration of the Competition, including but not limited to reading and learning from the Data, analyzing the Data, __modifying the Data__ and generally preparing your Submission and any underlying models and participating in forum discussions on the Website. ", "votes": 2 }, { "id": 180276, "postDate": "2017-05-04T17:58:26.520Z", "content": "Unless otherwise permitted by the terms of the Competition Website, Participants must use the Data solely for the purpose and duration of the Competition, including but not limited to reading and learning from the Data, analyzing the Data, modifying the Data and generally preparing your Submission and any underlying models and participating in forum discussions on the Website.
\n
As far as the passage you copied goes, this says you can only modify the data for the purpose and duration of the competition. This allows all of those things as long as it's within the comp
", "rawMarkdown": "As far as the passage you copied goes, this says you can only modify the data for the purpose and duration of the competition. This allows all of those things as long as it's within the comp", "votes": 1 } ] } ], "comments": [ { "id": 171108, "author_name": "Igor Barinov", "author_url": "", "post_date": "2017-03-28T17:32:15.603000", "content": "Is it for a model pretrained by me or using pretrained model by anyone, e.g. VGG?\nDo you have any requirement for the model: where to host the file, file format etc
", "votes": 3, "replies": [ { "id": 175750, "author_name": "Ricardo Luján", "author_url": "", "post_date": "2017-04-17T10:57:02.027000", "content": "I would suggest you to read the \"Pretrained models thread\" for other recent Kaggle competitions. It will solve the most of your doubts about how to proceed. You've got one here;\nhttps://www.kaggle.com/c/the-nature-conservancy-fisheries-monitoring/discussion/25428
", "votes": 0, "replies": [] }, { "id": 192695, "author_name": "Shujian Liu", "author_url": "", "post_date": "2017-06-14T13:42:10.360000", "content": "Hi, Igor, did you get an answer?
", "votes": 0, "replies": [] } ] }, { "id": 192791, "author_name": "Konstantin Lopukhin", "author_url": "", "post_date": "2017-06-14T20:06:44.290000", "content": "torchvision models pretrained on ImageNet: http://pytorch.org/docs/torchvision/models.html
", "votes": 1, "replies": [] }, { "id": 176391, "author_name": "MaciejOsowski", "author_url": "", "post_date": "2017-04-20T09:50:19.927000", "content": "I guess that in this competition neither VGG nor IMAGENET will help much. They just might with a lot of finetuning. The best idea I found so far is using nearest neighbors first on train pictures numbers ( as lions tend to group in same numbers on islands). Puting those in folders, training conv network in recognising them and testing. Does it sound sane? I guess dot approach will be neither doable nor any help for real life solution for NOAA
", "votes": 1, "replies": [] }, { "id": 176278, "author_name": "Liam Larsen", "author_url": "", "post_date": "2017-04-19T19:23:53.510000", "content": "Since the thread here is not quite understood. As an example you may want to use one of these pretrained models and tweak it to your desires.
\n\nkeras inception v3 pretrained on imagenet
", "votes": 0, "replies": [] }, { "id": 196256, "author_name": "LivingProgram", "author_url": "", "post_date": "2017-06-26T21:06:08.910000", "content": "keras pretrained resnet50 https://github.com/fchollet/deep-learning-models/releases/download/v0.2/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5
", "votes": 0, "replies": [] }, { "id": 195679, "author_name": "mokp", "author_url": "", "post_date": "2017-06-24T14:02:01.620000", "content": "", "votes": 0, "replies": [] }, { "id": 195166, "author_name": "jeffalltogether", "author_url": "", "post_date": "2017-06-22T21:50:51.487000", "content": "Faster RCNN pre-trained on VOC 2007 challenge data set
", "votes": 0, "replies": [] }, { "id": 195137, "author_name": "albu", "author_url": "", "post_date": "2017-06-22T20:12:23.490000", "content": "", "votes": 0, "replies": [] }, { "id": 194891, "author_name": "kuhung ", "author_url": "", "post_date": "2017-06-22T05:04:22.887000", "content": "Keras applications https://keras.io/applications/
", "votes": 0, "replies": [] }, { "id": 194872, "author_name": "", "author_url": "", "post_date": "2017-06-22T03:01:33.747000", "content": "", "votes": 0, "replies": [] }, { "id": 194814, "author_name": "Tuomas Tikkanen", "author_url": "", "post_date": "2017-06-21T21:19:40.933000", "content": "Keras VGG16 with weights pre-trained on ImageNet.
", "votes": 0, "replies": [] }, { "id": 194531, "author_name": "Vlad", "author_url": "", "post_date": "2017-06-20T23:41:43.847000", "content": "Initially I thought as long as someone posted the data source (imagenet, posted a couple of times) in here it's acceptable to use it without commenting here, but looking at the comments everyone seems to post what they use.
\n\nSo I think I just missed the deadline, but I was planning to use Inception v1 pretrained on ImageNet from here https://github.com/tensorflow/models/tree/master/slim (weights http://download.tensorflow.org/models/inception_v1_2016_08_28.tar.gz). Is this still acceptable? If not I just won't use it.
", "votes": 0, "replies": [] }, { "id": 194512, "author_name": "firolino", "author_url": "", "post_date": "2017-06-20T22:11:19.007000", "content": "Keras VGG16 pretrained on imagenet
", "votes": 0, "replies": [] }, { "id": 194508, "author_name": "Hiromichi NOMATA", "author_url": "", "post_date": "2017-06-20T21:57:06.820000", "content": "https://keras.io/applications/\nhttps://github.com/kentsommer/keras-inceptionV4\nhttps://gist.github.com/flyyufelix/7e2eafb149f72f4d38dd661882c554a6
", "votes": 0, "replies": [] }, { "id": 194448, "author_name": "Selim Seferbekov", "author_url": "", "post_date": "2017-06-20T16:33:51.020000", "content": "https://keras.io/applications/
\n\nhttps://github.com/rykov8/ssd_keras
", "votes": 0, "replies": [] }, { "id": 194438, "author_name": "Mohammad Azam Khan", "author_url": "", "post_date": "2017-06-20T15:56:38.323000", "content": "VGG and ResNet
", "votes": 0, "replies": [] }, { "id": 194436, "author_name": "Matt Boyd", "author_url": "", "post_date": "2017-06-20T15:49:24.780000", "content": "VGG16 Keras pretrained
", "votes": 0, "replies": [] }, { "id": 194412, "author_name": "", "author_url": "", "post_date": "2017-06-20T13:07:08.997000", "content": "https://github.com/tensorflow/models/tree/master/slim#Pretrained
\n\nhttps://keras.io/applications/
", "votes": 0, "replies": [] }, { "id": 194251, "author_name": "Nick Knyazev", "author_url": "", "post_date": "2017-06-19T21:24:09", "content": "", "votes": 0, "replies": [] }, { "id": 194130, "author_name": "amaia", "author_url": "", "post_date": "2017-06-19T12:49:41.733000", "content": "http://download.tensorflow.org/models/object_detection/ssd_inception_v2_coco_11_06_2017.tar.gz
\n\nhttp://download.tensorflow.org/models/object_detection/rfcn_resnet101_coco_11_06_2017.tar.gz
\n\nhttp://download.tensorflow.org/models/object_detection/faster_rcnn_resnet101_coco_11_06_2017.tar.gz
\n\n\n\nhttps://pjreddie.com/media/files/yolo-voc.weights
\n\nSSD500 ILSVRC\nhttps://drive.google.com/open?id=0BzKzrI_SkD1_X2ZCLVgwLTgzaTQ
\n\nSSD512 07++12++COCO\nhttps://drive.google.com/open?id=0BzKzrI_SkD1_NVVNdWdYNEh1WTA
\n\nhttp://cs.unc.edu/~wliu/projects/SSD/ZF_conv_reduced.caffemodel
", "votes": 0, "replies": [] }, { "id": 194085, "author_name": "gheeraej", "author_url": "", "post_date": "2017-06-19T10:21:53.500000", "content": "Hi, we are using pretrained VGG16
", "votes": 0, "replies": [] }, { "id": 194016, "author_name": "Alex P", "author_url": "", "post_date": "2017-06-19T05:50:51.670000", "content": "Keras VGG16 trained on Imagenet
", "votes": 0, "replies": [ { "id": 194018, "author_name": "Alex P", "author_url": "", "post_date": "2017-06-19T05:51:51.970000", "content": "Fast RCNN based on AlexNet trained on Imagenet
", "votes": 0, "replies": [] } ] }, { "id": 193947, "author_name": "Artem.Sanakoev", "author_url": "", "post_date": "2017-06-18T22:14:13.040000", "content": "Tensorflow slim pretrained models: https://github.com/tensorflow/models/tree/master/slim#Pretrained
", "votes": 0, "replies": [] }, { "id": 193933, "author_name": "bmci", "author_url": "", "post_date": "2017-06-18T21:14:48.107000", "content": "Will try:
\n\nKeras pretrained models : https://keras.io/applications/
\n\nTensorflow pretrained models: https://github.com/tensorflow/models/tree/master/slim#Pretrained
", "votes": 0, "replies": [] }, { "id": 193852, "author_name": "sleepywyn", "author_url": "", "post_date": "2017-06-18T12:49:27.607000", "content": "weights pretrained on IMAGENET. here is the link: https://pjreddie.com/media/files/darknet19_448.conv.23
", "votes": 0, "replies": [] }, { "id": 193620, "author_name": "Johan M", "author_url": "", "post_date": "2017-06-17T06:11:51.483000", "content": "Keras VGG16 trained on Imagenet (Keras pretrained)
", "votes": 0, "replies": [ { "id": 194017, "author_name": "Alex P", "author_url": "", "post_date": "2017-06-19T05:51:25.277000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 193224, "author_name": "Chan", "author_url": "", "post_date": "2017-06-15T20:36:06.857000", "content": "ResNet, Inception, Inception-ResNet
", "votes": 0, "replies": [] }, { "id": 192891, "author_name": "QuangTeo", "author_url": "", "post_date": "2017-06-15T03:25:12.187000", "content": "I'm using VGG-16 from https://keras.io/applications/
", "votes": 0, "replies": [] }, { "id": 192767, "author_name": "", "author_url": "", "post_date": "2017-06-14T18:17:48.530000", "content": "", "votes": 0, "replies": [] }, { "id": 192739, "author_name": "", "author_url": "", "post_date": "2017-06-14T16:31:43.190000", "content": "", "votes": 0, "replies": [] }, { "id": 192726, "author_name": "", "author_url": "", "post_date": "2017-06-14T15:24:13.760000", "content": "", "votes": 0, "replies": [] }, { "id": 192724, "author_name": "", "author_url": "", "post_date": "2017-06-14T15:18:24.810000", "content": "", "votes": 0, "replies": [] }, { "id": 192474, "author_name": "", "author_url": "", "post_date": "2017-06-13T16:47:38.513000", "content": "", "votes": 0, "replies": [] }, { "id": 192023, "author_name": "", "author_url": "", "post_date": "2017-06-12T13:38:37.250000", "content": "", "votes": 0, "replies": [] }, { "id": 191920, "author_name": "", "author_url": "", "post_date": "2017-06-12T05:26:24.070000", "content": "", "votes": 0, "replies": [] }, { "id": 191632, "author_name": "", "author_url": "", "post_date": "2017-06-11T06:23:41.510000", "content": "", "votes": 0, "replies": [] }, { "id": 191166, "author_name": "", "author_url": "", "post_date": "2017-06-09T14:00:03.070000", "content": "", "votes": 0, "replies": [] }, { "id": 190295, "author_name": "", "author_url": "", "post_date": "2017-06-07T13:00:36.687000", "content": "", "votes": 0, "replies": [] }, { "id": 189450, "author_name": "", "author_url": "", "post_date": "2017-06-06T04:06:43.743000", "content": "", "votes": 0, "replies": [] }, { "id": 188351, "author_name": "", "author_url": "", "post_date": "2017-06-02T17:58:38.960000", "content": "", "votes": 0, "replies": [] }, { "id": 188350, "author_name": "", "author_url": "", "post_date": "2017-06-02T17:56:44.770000", "content": "", "votes": 0, "replies": [] }, { "id": 184899, "author_name": "", "author_url": "", "post_date": "2017-05-23T13:10:22.480000", "content": "", "votes": 0, "replies": [] }, { "id": 183276, "author_name": "", "author_url": "", "post_date": "2017-05-17T16:15:19.097000", "content": "", "votes": 0, "replies": [] }, { "id": 183265, "author_name": "", "author_url": "", "post_date": "2017-05-17T15:33:54.180000", "content": "", "votes": 0, "replies": [] }, { "id": 182789, "author_name": "", "author_url": "", "post_date": "2017-05-15T21:11:44.587000", "content": "", "votes": 0, "replies": [] }, { "id": 182576, "author_name": "", "author_url": "", "post_date": "2017-05-14T17:38:37.180000", "content": "", "votes": 0, "replies": [] }, { "id": 182481, "author_name": "", "author_url": "", "post_date": "2017-05-14T03:41:30.090000", "content": "", "votes": 0, "replies": [] }, { "id": 182011, "author_name": "", "author_url": "", "post_date": "2017-05-11T23:53:11.500000", "content": "", "votes": 0, "replies": [ { "id": 192721, "author_name": "", "author_url": "", "post_date": "2017-06-14T15:06:58.773000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 180539, "author_name": "", "author_url": "", "post_date": "2017-05-05T19:51:43.940000", "content": "", "votes": 0, "replies": [] }, { "id": 180506, "author_name": "", "author_url": "", "post_date": "2017-05-05T18:01:16.377000", "content": "", "votes": 0, "replies": [] }, { "id": 180345, "author_name": "", "author_url": "", "post_date": "2017-05-05T02:04:55.747000", "content": "", "votes": 0, "replies": [] }, { "id": 180285, "author_name": "", "author_url": "", "post_date": "2017-05-04T18:45:42.720000", "content": "", "votes": 0, "replies": [ { "id": 180787, "author_name": "", "author_url": "", "post_date": "2017-05-07T07:30:50.270000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 180069, "author_name": "", "author_url": "", "post_date": "2017-05-04T00:10:28.220000", "content": "", "votes": 0, "replies": [] }, { "id": 171051, "author_name": "", "author_url": "", "post_date": "2017-03-28T13:39:52.107000", "content": "", "votes": 0, "replies": [] }, { "id": 193568, "author_name": "", "author_url": "", "post_date": "2017-06-16T23:24:31.277000", "content": "", "votes": 0, "replies": [] }, { "id": 179968, "author_name": "", "author_url": "", "post_date": "2017-05-03T13:59:43.097000", "content": "", "votes": 0, "replies": [ { "id": 180011, "author_name": "", "author_url": "", "post_date": "2017-05-03T17:35:04.880000", "content": "", "votes": 2, "replies": [] }, { "id": 180276, "author_name": "", "author_url": "", "post_date": "2017-05-04T17:58:26.520000", "content": "", "votes": 1, "replies": [] } ] } ], "raw_markdown_by_id": { "171108": "Is it for a model pretrained by me or using pretrained model by anyone, e.g. VGG?\nDo you have any requirement for the model: where to host the file, file format etc", "170868": "You should feel free to use pretrained models for your entries in this competition, but you must post them here before the competition's entry deadline.", "192791": "torchvision models pretrained on ImageNet: http://pytorch.org/docs/torchvision/models.html", "176391": "I guess that in this competition neither VGG nor IMAGENET will help much. They just might with a lot of finetuning. The best idea I found so far is using nearest neighbors first on train pictures numbers ( as lions tend to group in same numbers on islands). Puting those in folders, training conv network in recognising them and testing. Does it sound sane? I guess dot approach will be neither doable nor any help for real life solution for NOAA ", "176278": "Hello, everyone!\n----------------\n\n - [YOLO][1]\n - [Keras pretrained][2]\n - [Fast.ai][3]\n - [Imagenet][4]\n\n\n----------\n\n\nSince the thread here is not quite understood. As an example you may want to use one of these pretrained models and tweak it to your desires.\n\n\n----------\n\n\n [1]: https://pjreddie.com/darknet/yolo/\n [2]: https://keras.io/applications/#xception\n [3]: http://course.fast.ai/\n [4]: http://data.dmlc.ml/mxnet/models/imagenet/", "196570": "keras inception v3 pretrained on imagenet", "196256": "keras pretrained resnet50 https://github.com/fchollet/deep-learning-models/releases/download/v0.2/resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5", "195679": "[darknet19_448.weights][1]\n\n [1]: https://pjreddie.com/darknet/", "195166": "Faster RCNN pre-trained on VOC 2007 challenge data set", "195137": "[TF slim mobilenet][1]\n\n\n [1]: https://github.com/tensorflow/models/tree/master/slim#Pretrained", "194891": "Keras applications https://keras.io/applications/", "194872": "[keras application][1]\n\n\n [1]: https://keras.io/applications/", "194814": "[Keras VGG16](https://github.com/fchollet/keras/blob/master/keras/applications/vgg16.py) with weights pre-trained on ImageNet.", "194531": "Initially I thought as long as someone posted the data source (imagenet, posted a couple of times) in here it's acceptable to use it without commenting here, but looking at the comments everyone seems to post what they use.\n\nSo I think I just missed the deadline, but I was planning to use Inception v1 pretrained on ImageNet from here https://github.com/tensorflow/models/tree/master/slim (weights http://download.tensorflow.org/models/inception_v1_2016_08_28.tar.gz). Is this still acceptable? If not I just won't use it.", "194512": "Keras VGG16 pretrained on imagenet", "194508": "https://keras.io/applications/\nhttps://github.com/kentsommer/keras-inceptionV4\nhttps://gist.github.com/flyyufelix/7e2eafb149f72f4d38dd661882c554a6", "194448": "https://keras.io/applications/ \n\nhttps://github.com/rykov8/ssd_keras", "194438": "VGG and ResNet", "194436": "VGG16 Keras pretrained", "194412": "https://github.com/tensorflow/models/tree/master/slim#Pretrained\n\nhttps://keras.io/applications/", "194251": "[Keras pretrained][1] \n\n\n [1]: https://keras.io/applications/#xception", "194130": "http://download.tensorflow.org/models/object_detection/ssd_inception_v2_coco_11_06_2017.tar.gz\n\nhttp://download.tensorflow.org/models/object_detection/rfcn_resnet101_coco_11_06_2017.tar.gz\n\nhttp://download.tensorflow.org/models/object_detection/faster_rcnn_resnet101_coco_11_06_2017.tar.gz\n\nhttp://download.tensorflow.org/models/object_detection/faster_rcnn_inception_resnet_v2_atrous_coco_11_06_2017.tar.gz\n\nhttps://pjreddie.com/media/files/yolo-voc.weights\n\nSSD500 ILSVRC\nhttps://drive.google.com/open?id=0BzKzrI_SkD1_X2ZCLVgwLTgzaTQ\n\nSSD512 07++12++COCO\nhttps://drive.google.com/open?id=0BzKzrI_SkD1_NVVNdWdYNEh1WTA\n\nhttp://cs.unc.edu/~wliu/projects/SSD/ZF_conv_reduced.caffemodel", "194085": "Hi, we are using pretrained VGG16", "194016": "Keras VGG16 trained on Imagenet", "193947": "Tensorflow slim pretrained models: https://github.com/tensorflow/models/tree/master/slim#Pretrained", "193933": "Will try:\n\nKeras pretrained models : https://keras.io/applications/\n\nTensorflow pretrained models: https://github.com/tensorflow/models/tree/master/slim#Pretrained\n", "193852": "weights pretrained on IMAGENET. here is the link: https://pjreddie.com/media/files/darknet19_448.conv.23", "193620": "Keras VGG16 trained on Imagenet (Keras pretrained)", "193224": "ResNet, Inception, Inception-ResNet", "192891": "I'm using VGG-16 from https://keras.io/applications/", "192767": "VGG16 from https://keras.io/applications/", "192739": "I am using the pretrained models with imagenet weights from Keras: https://keras.io/applications/", "192726": "VGG16 and Resnet50", "192724": "I'm using keras and ssd https://github.com/rykov8/ssd_keras with the weights pretrained on VOC2007 (https://mega.nz/#F!7RowVLCL!q3cEVRK9jyOSB9el3SssIA)", "192474": "using keras pretrained models. \nAdditionally using mobilenet I pretrained on imagenet. https://github.com/danzelmo/mobilenet\n\nEdit: also https://github.com/tensorflow/models/tree/master/slim#Pretrained, and \nhttps://github.com/tensorflow/models/blob/master/object_detection/g3doc/detection_model_zoo.md", "192023": "Resnet50.https://github.com/tensorflow/models/tree/master/slim", "191920": "Hi, we are using Inception V3", "191632": "I'm using keras applications https://keras.io/applications/\nand ssd keras port https://github.com/rykov8/ssd_keras", "191166": "Inception-v4 from https://github.com/tensorflow/models/tree/master/slim", "190295": "I use torch ResNet.\nhttps://github.com/facebook/fb.resnet.torch/tree/master/pretrained\n", "189450": "Pretrained model: VGG16 trained on Imagenet.", "188351": "https://keras.io/applications/", "188350": "RESNET (-50, -101) pretrained on ImageNet", "184899": "Pretrained ZF", "183276": "https://keras.io/applications/ also", "183265": "Keras applications https://keras.io/applications/", "182789": "VGG16", "182576": "", "182481": "VGG16", "182011": "keras frcnn