Comparative_study

Comparative study for deep learning frameworks to test which framework works out to be the best for beginners.

##Project Description:

#Authors :

1. Aditya R. Karmarkar ([email protected])
2. Abdullah Alnajim ([email protected])
3. Abdulrahman Alshammari ([email protected])

Start date : 03/31/2017 End date : 05/05/2017 Frameworks to be studied :

1. Torch
2. Tensorflow
3. Caffe
4. deeplearning4j
5. Theano
6. CNTK

Fields to be considered under this study:

1. Computationl Biology
2. GeoPhysics
3. Statistics
4. Classical ML
5. Imaging

#Aim : Study deep learning frameworks and create guidelines for beginners to use such frameworks. Also, try to rate frameworks based on their usage, simplicity and avalability of documentation.

#Methodology :

1. Learn and sort framework according to popularity within the community
2. Pick up dataset to create models using frameworks (More the datasets, comparison can be distinct).
3. Try to generalize the models so that complexity of framework can be analysed.
4. Based on complexity and learning time spent on each framework can be plotted to draw picture of simplicity acroos framework learning.
5. Conclude

#Results: TBA

#Conclusion: TBA

Classified content******* Image dataset used for Object detection using DNN is copy-right content and ony used for research. If any researcher or Student want to study it then you can request for it here. http://www.image-net.org/request.

---

Project folder has following items:

1. Datasets --> All datasets used for comparative study. stored as .data{actual data} and .names{description of data}
2. Framework is directory which holds scripts written for all the frameworks and related files.
3. Guidelines --> holds guideline document of the study
4. Results --> holds all the plots, stats and other statistical notes