Deep Learning Based End-to-End System for Human Gait Recognition

Project Title

Deep Learning Based End-to-End System for Human Gait Recognition

Project Area of Specialization Artificial IntelligenceProject Summary

•Gait recognition is a computer vision technique that recognize the human as the way they walk.

•Unlike other biometric approaches such as a face identification and fingerprint recognition, gait recognition process can identifying a person’s gait from a distance, even in dark surroundings(unfavourable conditions).

•Gait recognition can be used in video surveillance to detect criminal activities, because it does not requires the subject’s interaction.

•It can also be used in medical fields to detect diseases/ disabilities by monitoring the patient’s gait.

•This project deals with recognition of individual with variations in dressing and object carring condition as these variations change the human gait.

•Our project uses Convolution Nerual Network for extracting features and then train SVM with these features and recognize gait based on these classification..

•This project is done on Matlab.

Project Objectives

•Improve precision of feature extrection.

•Achive high classification accuracy.

•Recognition in minimum time.

•Make Human Gait Recognition less resource heavy.

•Provide an End to End system for HGR.

Project Implementation Method

•Capture videos using CCTV.

•Preprocess videos.

•Split these videos into frames.

•Do a 30,70 split for traing and testing.

•Extract features using Alexnet,VGG16 and VGG19.

•Apply Generation learning Algorithims for refinment of features.

•Train SVM.

•Classification and Recognition.

•Compare and pridict.

•Testing.

•Generate result.

Benefits of the Project

•Main goal of this project is "Peace and Justice Strong Institutions".

•This system is useful for security and surveillance as it does not reqire any human involvement for its operation and can easily identify suspect.

•Gait recognition can be used in video surveillance to detect criminal activities, because it does not requires the subject’s interaction.

•Video surveillance is very useful for law enforcement as any misbehaviour can be spoted quickly.

•Any person wearing suspectible clothing can be identified e.g. a person wearing a jacket in summer or in restricted area or a person holding gun/ ammonation.

•It can also be used in medical fields to detect diseases(skeletal issues) by monitoring the patient’s gait.

•And can be helpful for find disabilities.

Technical Details of Final Deliverable

•Providing an image processing system that works on the basis of Deep learning and Artificial Intelligence. 

•Use different feature extrection algorithims e.g. Alexnet, VGG16, VGG19.

•Select a feature layer and do train SVM for classifitaion.

•Use these classification for recognition.

•Provide an End-to-End system for Hman Gait Recognition.

Final Deliverable of the Project Software SystemCore Industry SecurityOther Industries Medical Core Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Peace and Justice Strong InstitutionsRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	75065
GPU Nvidia GeForce GTX 1650	Equipment	2	27500	55000
CCTV HIKVISION 5 Megapixel	Equipment	1	4600	4600
Matlab 2019b with Deep Learning Toolbox(Student License)	Equipment	1	4475	4475
DDR-4 Ram(8 GB)	Equipment	1	4500	4500
Stationary	Miscellaneous	1	1540	1540
Printing of Proposal Reports	Miscellaneous	1	350	350
Printing of Panaflex	Miscellaneous	2	750	1500
Printing of Report	Miscellaneous	2	620	1240
Travel expenditures	Miscellaneous	1	1360	1360
Other Utilities	Miscellaneous	1	500	500