DeepHPD - Deep Human Presence Detection

Contributers : Sree Harsha Nelaturu, Anith Patel, Avyay Sah

DeepHPD is a novel application which uses Deep Learning in order to detect the presence of a human being. This is done by using a combination of Convolutional Neural Networks and OpenCV which along with a Raspberry pi and Basler Dart Camera, which blinks an LED when it detects human presence.

Software Used :

• Tensorflow
• OpenCV
• Numpy
• RPi-GPIO
• CUDA

Download the weights of the trained model and place them in the same folder

Weights.

The Tensorflow graph also to be placed in the same folder

Retrained Graph.

The two classes provided by the classifier are:

1. human
2. nohuman

For running the project, Key files required are :

1. Inference.py - Uses weights of trained Convolutional Neural Network for classification
2. LED-Blink.py - Script to be run to use inferred labels from footage in order to detect humans and blink LED
3. retrained_graph.pb - Contains Tensorflow static graph for computation
4. inception - Folder which contains the bottlenecks and weights for the moodel
5. retrained_labels.txt - Text file that contains the labels of the retrained model.

Once these resources are compiled, run LED-Blink.py to access the program.

Sample:

[INPUT]:

Footage via Basler DART is taken via OpenCV

[OUTPUT]:

Label : "human" LED Blinks