Custom_Object_Detection_tensorflow

• We are going to create a Deep Learning model using Tensorflow which can be trained to detect custom Object.
• We use Transfer Learning method.
• The Base Model we use is SSD MobileNet V2 FPNLite 320x320.
• The Base Model is available in Tensorflow Model Garden TF_MODEL_GARDEN

Prerequisite

• Python 3
• Tensorflow
• Anaconda or mini-conda
• Jupyter Notebook

Info

• The Over-all Steps are divided into two jupyter notebook
• The first notebook(Image Collection.ipynb) contains all the code which is used to collect the data for training and testing and image labeling.
• The second notebook(model.ipynb) contains all the code which deals with model creation/tweaking , evaluation and prediction.

Image Collection.ipynb

In this notebook we setup our starting folders and collect the data which we will use for model training and testing and also label the image using LabelImg.

model.ipynb

• In this notebook we first download and install the tensorflow object detection api from the tensorflow model garden.
• Then we create a label map for our model.
• Then we create Tfrecords using GenerateTFRecord.
• Now we import pipeline config file from the api folder and update it according to our custom objects/labels.
• Now we train and Evaluate our model .