A Python based real time object detection application utilizing SSD_MOBILENET algorithm and Tensorflow APIs, designed specifically for the blind victims.
- Real time interface
- Can be interacted with a laptop webcam and Android Mobile webcam as well
- Voice activated outputs
- Real time Distance estimation
- Warning alert system for closing objects
- OpenCv - Open Source Computer Vision Library for image and video processing.
- Pyttsx3 - Text-to-speech conversion library in Python, for voice genneration module
- Pytesseract - Python wrapper for Google's Tesseract-OCR Engine, allowing you to extract text from images
- Uses Pre-trained SSD detection model trained on COCO DATASETS
- Uses Resnet based architectural approach for feature extraction
- Implemented it by using TensorFlow Object Detection API
- Install all the required Libraries as per your system’s requirement. Visit Tensorflow for detailed installation analysis
- You can simply install it on Anaconda Prompt with following commands.
#for cpu
pip install tensorflow
#for gpu
pip install tensorflow-gpu
Download the TensorFlow model repository
-
Convert the .protos file to .py file extensions.
-
Head on to the Google Protobuf Releases
-
Download the protobuf version which satisfies your system compatibility.
-
Add it’s path in the environment variable
-
Inside of tensorflow/models/research/ directory hit the following command:
#From tensorflow/models/research/ protoc object_detection/protos/*.proto --python_out=.
- By the end of this you will have successfully converted your protos file into python files.
- You can use any one of the pre-trained models provided by Tensorflow depending upon your system specifications from here
- For a faster accuracy you can go with SSD DETECTION and for better accuracy you can go with MASK RCNN
- Siince most of the system shows smooth performance with SSD Mobile_Net DETECTION, we will be using that.
- This model is based on the ideology of THE MobileNet model
- Fork the repository at your profile
- Git Clone the repository to your local machine.
- pip install - r requirements
- Run webcam_blind_voice.py
-
Utilizes values of average real height of the object(mm), focal length(mm) and sensor height(mmm) of the camera, image height in pixels.
-
Formula:
distance_apx=round(((f_mm*real_height_mm*img_height_px)/(object_height_px*sensor_height_mm)))