Skip to content

Final project for Nile university CIT-651 Machine learning class - Fall 19

License

Notifications You must be signed in to change notification settings

MahmoudTamam/Human-activity-Recoginition

Repository files navigation

Human activity recoginition with Deep learning

Final project for Nile university CIT-651 Machine learning class - Fall 19 Experiments are performed on Pascal VOC 2012, Kaggle statefarm distracted driver and UCI Har dataset.

UCI HAR Dataset

Fully Connected network

Trained on Raw data features 561x1. Image description Image description

1D CNN

Trained with time stamp readings 9x128. Image description

VOC 2012 Dataset (Indoor, Outdoor, ... )

Trained with 512x512 images with/without bounding box cropping with random augmentation. Image description Image description

Statefarm dataset (Inside vehicle)

Trained with 480x640 images with random augmentation. Image description Image description

References:

Template structure and config scripts from Template link