Aco Shopov Real Time Sign Language Detection with Tensorflow Object Detection and Python and Deep Learning
Using Tensorflow Object Detection API, Python, pre-trained model SSD MobNet V2, created datasets (labeled with LabelImg). Includes detection of 7 gestures : Da, Nadevanja, Stemneto, Sonce, Sili, Istorija, Pet i Sedumnaeset
- path_names.py - Includes paths of whole directory for easier usage in code
- collect_images.py - Collects images from web-cam and creates data set, then you label images with labelImg, and divide them in train and test set
- label_map.py - Creates label map
- generate_tfrecord.py - Generates TF records (train.record and test.record)
- update_config_for_learning.py - Updates pre-trained pipeline config to our parameters
- load_model.py - Loads specified model (Includes checkpoints from 1k to 10k epochs of trained model)
- test_picture.py - Predicts image from test set
- test_video.py - Main script that predicts Sign Language in real time
- text files - include all instructions and commands to train model, evaluate model, and create the TF Records
Demo Video : https://www.youtube.com/watch?v=HuW4YDxdDJE&ab_channel=NikolaDinevski
-Petar Atanasovski -Nikola Dinevski