ENHANCED PROTECTION
USING CNN
Introduction
In this work, we propose the use of an existing pre-trained 3D
Convolutional Neural Network (CNN), named C3D . Here we use
CNN for violence detection in videos through cameras and going
to implement next move to protect us .
3D CNN can understand the movement over time in
videos. It does this by using 3D convolution and
pooling. This means it looks at groups of frames
stacked together to figure out how things change
over time
3D CNN :
Figure of Convolution and pooling
Figure of frame detection in videos
VIOLENCE DETECTION :
Detecting violence in video data presents a significant challenge due
to the intricate identification of complex sequential visual patterns.
There are many unimportant frames so it takes more memory .so , we
first detected the persons in the video stream using a pre-trained
CNN model.
Only the sequence of 16 frames containing persons was passed to the
3D CNN model for final prediction, which helped achieve effiecient
processing.
ABNORMAL ACTIVITY: NORMAL ACTIVITY:
OUR IDEA OF IMPLEMENTATION :
 It is impossible to detect every violence through cctv camers
Accurately.
 By implementing our idea through cam specs we can able
detect every video frames accurately using 3D CNN.
 And it’s easy to protect us from major threads like harassment
And informal attacks.
 Hope ,Even it works better than Kavalan SOS which is implemented by
our TN government
DATA THROUGH
FRONT END:
Thanks!
PRESENTORS:
NAWFAL ARSATH M
PRAKASH RAJ S

Convolutional Neural Networks cnn pre.pptx

  • 1.
  • 2.
    Introduction In this work,we propose the use of an existing pre-trained 3D Convolutional Neural Network (CNN), named C3D . Here we use CNN for violence detection in videos through cameras and going to implement next move to protect us .
  • 3.
    3D CNN canunderstand the movement over time in videos. It does this by using 3D convolution and pooling. This means it looks at groups of frames stacked together to figure out how things change over time 3D CNN :
  • 4.
    Figure of Convolutionand pooling Figure of frame detection in videos
  • 5.
    VIOLENCE DETECTION : Detectingviolence in video data presents a significant challenge due to the intricate identification of complex sequential visual patterns. There are many unimportant frames so it takes more memory .so , we first detected the persons in the video stream using a pre-trained CNN model. Only the sequence of 16 frames containing persons was passed to the 3D CNN model for final prediction, which helped achieve effiecient processing.
  • 6.
  • 7.
    OUR IDEA OFIMPLEMENTATION :  It is impossible to detect every violence through cctv camers Accurately.  By implementing our idea through cam specs we can able detect every video frames accurately using 3D CNN.  And it’s easy to protect us from major threads like harassment And informal attacks.  Hope ,Even it works better than Kavalan SOS which is implemented by our TN government
  • 9.
  • 10.