You are on page 1of 13

SCENE CLASSCIFICATION USING

DEEP LEARNING

BY M. WASIF ASRAR
STUDENT NO:1810140
OUTLINE

1. Introduction
2. Aim and Objective
3. Summary of LR
4. Methodology
5. Software Implementation
6. Result
7. Discussion
8. Conclusion
9. Limitation
10.Bibliography
INTRODUCTION

When a scene is presented to humans, they are able to quickly identify the
scene or object, i.e., within a short period of exposure, by using this
concept absorb the incredible capability of the human vision system into
their algorithms for image processing(Zhang, 2016).

For example If the criminal is roaming around in the mall or in the middle
of the city to commit a crime, if it is recognized by a human, it take time to
respond and inform the authorities. Using scene classification with deep
learning and computer vision, it will be detected without any delay.
AIM

The aim of the project is to load, train and test different categories of images
using deep learning in matlab and to obtain result accuracy on test images
using SVM classifier for scene classification.
OBJECTIVES
• To load different types of images in matlab
• To train by extracting feature for scene classification
• To test images for scene classification
SUMMARY OF LR

What is AI ?

What is Machine Learning ?

What is deep learning?


BLOCK DIAGRAM
RESULT

• The images were categorizing when the data was uploaded in matlab. The bag of
feature method was used to extract the different feature from different images.
• In the classification learner, using the SVM classifier the prediction of the images
was modelled in confusion matrix and accuracy 71.9% were obtained.
DISCUSSION
CONCLUSION

• The matlab was able to classify the scene with different test image categories.
• The features of the different images were extracted by the method bag of feature.
• 71.9% accuracy was obtained by support vector machine algorithm classifier in the
classification app of matlab.
• Prediction of all data image categories were taken place in the confusion matrix,
where it classifies all 4 classes of different image categories and how many scenes
are predicted to each other due to similar features
LIMITATION

The biggest limitation of deep learning models is that they learn through
observations. This means they only know what was in the data they trained on. If
a user has a small amount of data or it comes from one specific source that is
not necessarily representative.
BIBLIOGRAPHY
Srivastava, Nitish, et al. "Dropout: a simple way to prevent neural networks from
overfitting." Journal of machine learning research (2014)

Bergstra, James, and Yoshua Bengio. "Random search for hyper-parameter optimization." Journal
of Machine Learning Research, Feb (2012)

Kim, Y. “Convolutional Neural Networks for Sentence Classification”, EMNLP (2014)

Severyn, Aliaksei, and Alessandro Moschitti. "UNITN: Training Deep Convolutional Neural Network
for Twitter Sentiment Classification." SemEval@ NAACL-HLT (2015)

Cho, Kyunghyun, et al. "Learning phrase representations using RNN encoder-decoder for
statistical machine translation." EMNLP (2014)

Ilya Sutskever et al. “Sequence to sequence learning with neural networks.” NIPS (2014)

Bahdanau et al. "Neural machine translation by jointly learning to align and translate." ICLR
(2015)
BIBLIOGRAPHY

• Zhang, L., Yang, F., Zhang, Y.D. and Zhu, Y.J., 2016, September. Road crack detection using deep convolutional
neural network. In 2016 IEEE international conference on image processing (ICIP) (pp. 3708-3712). IEEE.
• Gal, Y., Islam, R., Ghahramani, Z. “Deep Bayesian Active Learning with Image Data.” ICML (2017)
• Nair, V., Hinton, G.E. “Rectified linear units improve restricted boltzmann machines.” ICML (2010)
• Ronan Collobert, et al. “Natural language processing (almost) from scratch.” JMLR (2011)
• Kumar, Shantanu. "A Survey of Deep Learning Methods for Relation Extraction." arXiv preprint
arXiv:1705.03645 (2017)
• Li X, Shi J, Dong Y, Tao D. A survey on scene image classi cation. Scientia Sinica Informationis. 2015 Jul
10;45(7):827.
THANK YOU

You might also like