MrinalJain17 / Human-Activity-Recognition

Recognizing human activities using Deep Learning

Machine Learning Engineer Nanodegree

⚠️ This project is not maintained anymore ⚠️

Capstone Project - Human Activity Recognition

Recognizing human activities using Deep Learning

View the project notebook here - Link to Jupyter Notebook

Dataset

Recognition of Human Actions

• Walking.zip
• Jogging.zip
• Running.zip
• Boxing.zip
• Handwaving.zip
• Handclapping.zip

There are a total of 599 videos, with each category having 100 videos (with the exception of Handclapping having 99 videos).

All the videos were captured at 25fps frame rate. Each video has a spatial resolution of 160x120 pixels.

Instructions

1. Clone the repository and navigate to the downloaded folder.

   	git clone https://github.com/MrinalJain17/Human-Activity-Recognition.git
   	cd Human-Activity-Recognition
   

2. Unzip the compressed data files and store in the format as mentioned here

   • Use the helper function download_files() present in data_utils.py as follows to do this in your current working directory automatically. (The function will delete the compressed files after they are successfully extracted)

   	import data_utils
   	
   	data_utils.download_files()

3. The following file is corrupted which gives an error when being loaded. Delete it before proceeding.

   'person01_boxing_d4_uncomp.avi' (present in Data/Boxing/)

4. In order to read the videos, there is a helper class Videos in utils.py.

   	import numpy as np
   	from utils import Videos
   	
   	reader = Videos(target_size=(128, 128), 
   			to_gray=True, 
               		max_frames=40, 
               		extract_frames='first', 
               		required_fps=5, 
               		normalize_pixels=(-1, 1))
   	
   	videos = reader.read_videos(video_absolute_paths)

   Refer the code for a detailed documentation.
   This utility is being maintained in a seperate repository here

5. Run the following command to view the project notebook:

   	jupyter notebook human_activity_recognition.ipynb
   

Requirements

Python 3.x (preferably from the Anaconda Distribution)

Install FFmpeg on your machine

For Linux:

	$ sudo apt-get update
	$ sudo apt-get install libav-tools

For Windows or MAC/OSX:
Download the required binaries from here. Extract the zip file and add the location of binaries to the PATH variable

Additional Libraries:

• Scikit-video

   	pip install sk-video
  

• Tensorflow

   	pip install tensorflow
  

  For GPU support or a custom installation, follow the instructions given on the Tensorflow website.

• Keras

   	pip install keras
  

• tqdm - Required for displaying the progress bar.

   	pip install tqdm
  

These libraries will be required for successful execution of the project files.