Setting-up Tensorflow Object Detection API

Prerequisites:

• Tensorflow 1.4 or newer [pip install tensorflow-gpu]
• LabelImage [https://github.com/tzutalin/labelImg]
• pandas [pip install pandas]
• matplotlib [pip install matplotlib]

Get the code

Clone objectdetect repository from https://github.com/dotannn/objectdetect go to utils directory and run download_model sctipt :

python download_model.py --model <model_name>

if you don't know which model do you want, go to https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md to see a list of optional pre-trained models. or just leave it empty for default model.

Now, test the code by running run_inference.py script. this will run the pretrained model on the sample image in data/sample dir. here's the result:

Prepare data for training

Run training and debug results

Test on new data