Neural Networks: Types and Applications
Frameworks and datasets
Several DL frameworks and datasets have been developed in the last few years. various frameworks and libraries have also been used in order to expedite the work with good results. Through their use, the training process has become easier. Table 4 lists the most utilized frameworks and libraries.
Table 4 List of the most common frameworks and libraries
|
Framework |
License |
Core language |
Year of release |
Homepages |
|---|---|---|---|---|
|
TensorFlow |
Apache 2.0 |
C++ & Python |
2015 |
https://www.tensorflow.org/ |
|
Keras |
MIT |
Python |
2015 |
https://keras.io/ |
|
Caffe |
BSD |
C++ |
2015 |
http://caffe.berkeleyvision.org/ |
|
MatConvNet |
Oxford |
MATLAB |
2014 |
http://www.vlfeat.org/matconvnet/ |
|
MXNet |
Apache 2.0 |
C++ |
2015 |
https://github.com/dmlc/mxnet |
|
CNTK |
MIT |
C++ |
2016 |
https://github.com/Microsoft/CNTK |
|
Theano |
BSD |
Python |
2008 |
http://deeplearning.net/software/theano/ |
|
Torch |
BSD |
C & Lua |
2002 |
http://torch.ch/ |
|
DL4j |
Apache 2.0 |
Java |
2014 |
https://deeplearning4j.org/ |
|
Gluon |
AWS Microsoft |
C++ |
2017 |
https://github.com/gluon-api/gluon-api/ |
|
OpenDeep |
MIT |
Python |
2017 |
http://www.opendeep.org/ |
Based on the star ratings on Github, as well as our own background in the field, TensorFlow is deemed the most effective and easy to use. It has the ability to work on several platforms. (Github is one of the biggest software hosting sites, while Github stars refer to how well-regarded a project is on the site). Moreover, there are several other benchmark datasets employed for different DL tasks. Some of these are listed in Table 5.
Table 5 Benchmark datasets
|
Dataset |
Num. of classes |
Applications |
Link to dataset |
|---|---|---|---|
|
ImageNet |
1000 |
Image classification, object localization, object detection, etc. |
http://www.image-net.org/ |
|
CIFAR10/100 |
10/100 |
Image classification |
https://www.cs.toronto.edu/~kriz/cifar.html |
|
MNIST |
10 |
Classification of handwritten digits |
http://yann.lecun.com/exdb/mnist/ |
|
Pascal VOC |
20 |
Image classification, segmentation, object detection |
http://host.robots.ox.ac.uk/pascal/VOC/voc2012/ |
|
Microsoft COCO |
80 |
Object detection, semantic segmentation |
https://cocodataset.org/#home |
|
YFCC100M |
8M |
Video and image understanding |
http://projects.dfki.unikl.de/yfcc100m/ |
|
YouTube-8M |
4716 |
Video classification |
https://research.google.com/youtube8m/ |
|
UCF-101 |
101 |
Human action detection |
https://www.crcv.ucf.edu/data/UCF101.php |
|
Kinetics |
400 |
Human action detection |
https://deepmind.com/research/open-source/kinetics |
|
Google Open Images |
350 |
Image classification, segmentation, object detection |
https://storage.googleapis.com/openimages/web/index.html |
|
CalTech101 |
101 |
Classification |
http://www.vision.caltech.edu/Image_Datasets/Caltech101/ |
|
Labeled Faces in the Wild |
– |
Face recognition |
http://vis-www.cs.umass.edu/lfw/ |
|
MIT-67 scene dataset |
67 |
Indoor scene recognition |
http://web.mit.edu/torralba/www/indoor.htm |