# Official External Data Thread **URL:** \ **Category:** N+1 Fish, N+2 Fish\ **Created:** [September 13, 2017, 11:53pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434 "2017-09-13T23:53:39Z")\ **Posts on this page:** 15\ **Page:** 1
**Author:** ![bull](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/bull/32/8_2.png) [@bull](https://community.drivendata.org/u/bull)\ **Post date:** [September 13, 2017, 11:53pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/1 "2017-09-13T23:53:39Z")
If you use external data, [per this announcement](https://www.drivendata.org/competitions/48/identify-fish-challenge/announcements/), include a link to the data here! It must be freely publicly available. ---
**Author:** ![bull](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/bull/32/8_2.png) [@bull](https://community.drivendata.org/u/bull)\ **Post date:** [September 13, 2017, 11:57pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/2 "2017-09-13T23:57:17Z")
---
**Author:** ![ZFTurbo](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/zfturbo/32/580_2.png) [@ZFTurbo](https://community.drivendata.org/u/ZFTurbo)\ **Post date:** [September 26, 2017, 4:45pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/3 "2017-09-26T16:45:08Z")
I plan to use the following pretrained models: DenseNet: https ://github.com/flyyufelix/DenseNet-Keras ResNet-101: https ://gist.github.com/flyyufelix/65018873f8cb2bbe95f429c474aa1294 ResNet-152: https ://gist.github.com/flyyufelix/7e2eafb149f72f4d38dd661882c554a6 SqueezeNet: https ://github.com/rcmalli/keras-squeezenet Inception v4: https ://github.com/titu1994/Inception-v4/releases VGG16: https ://gist.github.com/baraldilorenzo/07d7802847aaad0a35d3 VGG19: https ://gist.github.com/baraldilorenzo/8d096f48a1be4a2d660d Other Keras models: https ://keras.io/applications/ ---
**Author:** ![lopuhin](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/lopuhin/32/532_2.png) [@lopuhin](https://community.drivendata.org/u/lopuhin)\ **Post date:** [October 1, 2017, 11:13am UTC](https://community.drivendata.org/t/official-external-data-thread/1434/4 "2017-10-01T11:13:00Z")
I plan to use pre-trained VGG-16 referenced here [https://github.com/amdegroot/ssd.pytorch#training-ssd](https://github.com/amdegroot/ssd.pytorch#training-ssd) and pre-trained pytorch models [http://pytorch.org/docs/master/torchvision/models.html](http://pytorch.org/docs/master/torchvision/models.html) ---
**Author:** ![saket](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/saket/32/559_2.png) [@saket](https://community.drivendata.org/u/saket)\ **Post date:** [October 7, 2017, 2:35pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/5 "2017-10-07T14:35:35Z")
Using keras weights from https ://keras.io/applications/ ---
**Author:** ![antor](https://avatars.discourse-cdn.com/v4/letter/a/eada6e/32.png) [@antor](https://community.drivendata.org/u/antor)\ **Post date:** [October 9, 2017, 8:04am UTC](https://community.drivendata.org/t/official-external-data-thread/1434/6 "2017-10-09T08:04:09Z")
> **[pytorch/vision](https://github.com/pytorch/vision/tree/master/torchvision/models)** > > vision - Datasets, Transforms and Models specific to Computer Vision ---
**Author:** ![dmytro](https://avatars.discourse-cdn.com/v4/letter/d/9fc348/32.png) [@dmytro](https://community.drivendata.org/u/dmytro)\ **Post date:** [October 9, 2017, 12:30pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/7 "2017-10-09T12:30:24Z")
In addition to listed pretrained models: DenseNet: https ://github.com/flyyufelix/DenseNet-Keras and other Keras models: https ://keras.io/applications/ I’m also using > **[rykov8/ssd\_keras](https://github.com/rykov8/ssd_keras)** > > ssd\_keras - Port of Single Shot MultiBox Detector to Keras ---
**Author:** ![sueiras](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/sueiras/32/572_2.png) [@sueiras](https://community.drivendata.org/u/sueiras)\ **Post date:** [October 11, 2017, 6:19pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/8 "2017-10-11T18:19:11Z")
Using this: [https://github.com/tensorflow/models/tree/master/research/object\_detection](https://github.com/tensorflow/models/tree/master/research/object_detection) ---
**Author:** ![alexparinov](https://avatars.discourse-cdn.com/v4/letter/a/ccd318/32.png) [@alexparinov](https://community.drivendata.org/u/alexparinov)\ **Post date:** [October 12, 2017, 12:08pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/9 "2017-10-12T12:08:51Z")
I am using pretrained pytorch models from [https://github.com/pytorch/vision](https://github.com/pytorch/vision) ---
**Author:** ![ezetl](https://avatars.discourse-cdn.com/v4/letter/e/e95f7d/32.png) [@ezetl](https://community.drivendata.org/u/ezetl)\ **Post date:** [October 13, 2017, 1:47am UTC](https://community.drivendata.org/t/official-external-data-thread/1434/10 "2017-10-13T01:47:51Z")
I’m planning to use this [https://github.com/yhenon/keras-frcnn](https://github.com/yhenon/keras-frcnn) And the models referenced here: [https://github.com/fchollet/deep-learning-models](https://github.com/fchollet/deep-learning-models) ---
**Author:** ![lonestranger](https://avatars.discourse-cdn.com/v4/letter/l/22d042/32.png) [@lonestranger](https://community.drivendata.org/u/lonestranger)\ **Post date:** [October 22, 2017, 10:12pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/11 "2017-10-22T22:12:05Z")
I am using pretrained ‘imagenet’ weights for VGG-16 from KERAS applications here: [https://keras.io/applications/#vgg16](https://keras.io/applications/#vgg16) ---
**Author:** ![harshml](https://avatars.discourse-cdn.com/v4/letter/h/13edae/32.png) [@harshml](https://community.drivendata.org/u/harshml)\ **Post date:** [October 26, 2017, 6:42pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/12 "2017-10-26T18:42:06Z")
Trying to use: [vgg1](https://gist.github.com/ksimonyan/211839e770f7b538e2d8#file-readme-md), [vgg2](https://gist.github.com/weiliu89/2ed6e13bfd5b57cf81d6), [mobilenet](https://github.com/shicai/MobileNet-Caffe), [resnet](https://github.com/cvjena/cnn-models/tree/master/ResNet_preact/ResNet10_cvgj), [pva](https://github.com/sanghoon/pva-faster-rcnn), [densenet](https://drive.google.com/open?id=0B7ubpZO7HnlCV3pud2oyR3lNMWs). ---
**Author:** ![HenriEps](https://avatars.discourse-cdn.com/v4/letter/h/e36b37/32.png) [@HenriEps](https://community.drivendata.org/u/HenriEps)\ **Post date:** [October 28, 2017, 9:56am UTC](https://community.drivendata.org/t/official-external-data-thread/1434/13 "2017-10-28T09:56:36Z")
I am also using pytorch/vision models. ---
**Author:** ![beluga](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/beluga/32/343_2.png) [@beluga](https://community.drivendata.org/u/beluga)\ **Post date:** [October 28, 2017, 1:25pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/14 "2017-10-28T13:25:32Z")
Pretrained keras weights from https ://keras.io/applications/ ---
**Author:** ![Daniel\_FG](https://avatars.discourse-cdn.com/v4/letter/d/e9a140/32.png) [@Daniel\_FG](https://community.drivendata.org/u/Daniel_FG)\ **Post date:** [October 29, 2017, 8:58pm UTC](https://community.drivendata.org/t/official-external-data-thread/1434/15 "2017-10-29T20:58:15Z")
Should we post the pretrained models we are using, even if they were already posted? Just in case, I’ve tried some of the pytorch models already mentioned.