Suspicious-Activity-Detection-from-videos

Extract frames from the video:

get_frames_from_video.ipynb - extract frames to their respective dir from video dataset.

Way1 Plain CNN

ImageclassifierCNN.ipynb- update required.. tried with CNN using keras/tf and a image as a single data of either normal or abnormal.
suspiciousDetectionFastai.ipynb - Used CNN to do the work using a frame as a single data point and used FastAI to do the task.

Way2 CNN + RNN

After getting frames:
Small_frames_to_feature_map.ipynb - you can store the high level featuremap of the frames in npy file through directory of frames, but just done to small dataset.
big_frames_to_feature_map.ipynb - just bigger version of Small_frames_to_feature_map.ipynb.

After getting npy file:

Small_RNNImageClassifier.ipynb - Predict from the numpy file of small dataset as dummy dataset of around 15 GB. RNNImageClassifier.ipynb - Predict from the numpy file of dataset using LSTM.

Other files

1.csv
2.csv
3.csv
4.csv
final.csv
These CSV files helps us to get frames of the video in correct order easily.

completing soon