Demo about anomalydetectioncvpr2018

[kjsr7]

1. Extract the features for every 16 frames for each video using the C3D model.
2. Use the save32_segment_features script of this repository to save the extracted features of each video into 32 segments.
3. Use training_anomaly_detector to train the model using your features. The output of training process would be model.json and weights of the neural network.
4. You can run Demo_GUI.py if you paste the corresponding videos in the Sample Videos folder. Demo_GUI.py shows you a graph with spikes if there exists an anomaly in the video.

[rajanjitenpatel]

@Shiny128
I have created C3D feature extraction using Google Colab for particular to this project. you can check it out.

I hope it will be helpful to you.

[LakyTT]

Demo_GUI.py is the Demo

[Shiny128]

What are the sequence of steps to be followed before running Demo_GUI.py ??
Could you please help me with the detailed description of the execution steps to get the right output?

[kjsr7]

Hi @Shiny128,
Have you found out the sequence of execution of files?

[Shiny128]

No I don't know the sequence of execution of files
Please guide me through it, if you are aware of the steps.

[kjsr7]

Before running demo_gui.py, you need to have the videos corresponding to the feature files present in the 'sample videos' directory of this repository. You need to place those videos in that directory. And then you can run the demo_gui.py

[Shiny128]

Hello @kjsr7
Please give me step by step execution of this project starting from feature extraction in simple words as I am new to programming.
The readme file does not have sufficient steps in detail to obtain final output
Please Help me

[Malathi15]

@kjsr7 Hi, SampleVideos folder contains .txt files. How can I create .txt files for my own test video file using demo_GUI.py?

[Shiny128]

Hello @kjsr7
I was unable to extract features using the C3D model using facebook/C3D code.
I am not able to proceed further with caffee installation
Please guide me through the steps that u followed to do the same in detail
Thank you

[WaqasSultani]

Hi Shiny
You don't have to use C3D. It's just the video features that we used in our paper. You can use I3D (https://github.com/deepmind/kinetics-i3d/blob/master/i3d.py) or Multifiber network (https://github.com/cypw/PyTorch-MFNet). I believe you should be able to compute them.
I got an email from someone that he used I3D instead of C3D and got better results.
I believe I3D or Multifiber network could be easily run on freely available GPU (Google Collab).

Best,

[slighting666]

Hi, I am a deep learning novice, can you tell me the sequence of code implementation in detail?@Shiny128