Hello, this is my first try to make something generalised, with the Python code in this repository you can develop an Object Detection model on your custom dataset [must be annotated, you can use labelImg to get Pascal VOC annotation], train it and test it. This work is based on Tensorflow, and it's library TFLITE-MODEL-MAKER, you can watch my explaination on this repository here.
- Introduction
- Requirements
- Usage
This is a simple Object detection wraper around Tensorflow Lite Model Maker, you can read more about it here.
With the help of this repository you can train an Object Detection model and save it in .tflite format.
Here are some basic requirements...
Python 3.8 or greater
Python packages Tensorflow Lite Model Maker and Pillow
there is a
requirements.txtfile that you can use to install the required packages
There is a config.py file in the repository, make sure to change that according to the requirements.
There are two main part of the project...
-
training a model
-
testing on a single file
before running any file make sure you take a look at the
config.pyfile and change the variables according to your need.
the dataset must have separate training and validation folders
make sure you go thorugh the config.py file, then run the file train_model.py, this will train the model and evaluate the model on the validation dataset and create a model in model folder
provide the path of the test image in the test.py folder to the variable INPUT_IMAGE_PATH and then run the file.
this will create a result folder and inside it, two files, one input.jpeg and annotated output.jpeg
If this repository is useful to you, please consider giving a start to it.
You can contact me to build any kind of Chatbot/AI/ML work.
Enjoy the life, Feel the music. Peace.