Deep Learning Starting Points
This article has some tips on where to start for reading up on deep learning using convolutional neural networks.
Deep Learning Starting Points

As is the case with most research areas, the most influential papers come from a wide variety of sources. There is no single definitive source for the top research.

This short article is intended for students who have a computer science background and are becoming interested in the subject of deep learning.

Some recommended publications are

CVPR - Computer Vision and Pattern Recognition Conference

ICCV - International Conference on Computer Vision

NIPS - Neural Information Processing Systems

MM - ACM International Conference on Multimedia

MMM - (ACM/IEEE) Multimedia Modeling Conference

Neural Networks - Journal on Neural Networks

Neurocomputing - Journal on Neural Computation

IEEE Trans MM - IEEE Transactions on Multimedia

IJCV - International Journal of Computer Vision

Also, many research reports are posted to, however, keep in mind that does not do peer review.

Many papers are easily accessible using Google Scholar ( - just paste in the title of the article and you will usually immediately find it. Note that if you are at a Leiden University or LIACS IP address (e.g. LIACS wireless), then you should have the option to download all papers from ACM (, IEEE Computer Society (, Springer (, and Elsevier ( for free.

If the paper is not downloadable from Google Scholar or any of the above, it is also standard practice to contact the first author of the paper and request a copy. They will usually be happy to send you a copy.

I suggest getting started by reading up on the surveys such as

"Deep learning in neural networks: An overview", in Neural Networks, 2015


"Deep learning for visual understanding: A review", in Neurocomputing, 2016

and interesting benchmarks such as

"Imagenet large scale visual recognition challenge", IJCV, 2015

and also neural programming frameworks such as Caffe:

Convolutional architecture for fast feature embedding", ACM MM, 2014

and TensorFlow:

TensorFlow: A System for Large-Scale Machine Learning", OSDI, 2016

and the top deep learning approaches from the ImageNet competition such as AlexNet:

"ImageNet Classification with Deep Convolutional Neural Networks", NIPS, 2012.

and VGG:

"Very deep convolutional networks for large-scale image recognition", Arxiv, 2014

and Residual Networks:

"Deep residual learning for image recognition", CVPR, 2016

Other interesting deep networks are

Dense Nets:

"Densely Connected Convolutional Networks", CVPR, 2017

and Convolutional Fusion Networks (CFN)

"On the Exploration of Convolutional Fusion Networks for Visual Recognition", MMM, 2017

In most cases, the authors provide the actual network for download also.

Please be aware that the learning curve is rather steep for deep learning. The training process can be quite tricky and it may be unclear or incomplete from the research paper. There are numerous things the authors may do to maximize the accuracy which may have nothing to do with the actual CNN. Your mileage may vary...

Good luck!

Dr. Michael S. Lew

Media Lab Overview
LIACS Homepage
MM Conf
ACM Multimedia
Science Direct
IEEE Library
LIACS Publications
ACM Digital Library