This System implements a CNN created using TensorFlow and uses OpenCV for utilities and Computer Vision and Image Processing tasks.
Variable size of images to be transformed into a uniform size through preprocessing utility-"cvt_dataset_into_bin" by changing sizes. 64 is preferable for quicker training.
- Clone the repository
- Make a dataset by capturing images of each sign and placing in Dataset/{letter of the sign}/{image_name.jpg}
- run train.py
- run GUI.py
- Ensemble of models for better prediction
- Using a RCNN or better models to look at "flowing" gesture indicating words or sentences in certain languages