Object Detection Python Examples

Run your models trained using Cloud Annotations with python.

Tflite Object Detection

Currently, python-tflite.py supports using Mobilenet-V1 SSD models trained using Cloud Annotations.

Output:

Summary Information

Image 7 of 9.
Inference time: 0.15027356147766113
----------
Inference Summary:
Highest Score: 0.9407029747962952
Highest Scoring Box: [0.60926155 0.47011317 0.67576766 0.56898813]
----------
Image shape: (563, 1000, 3)
Boxes shape: (1917, 4)
Classes shape: (1917,)
Scores shape: (1917,)
['plate: 94%']
Image Saved
==========

Saved Image:

Perform object detection with your model

Note: to find a list of all models trained do:

cacli list

To use a custom model, perform

cacli download <model_name>

For example, if the downloaded files were saved to /path/to/<model_name> :

Our tflite model is stored in <model_name>/model_android/model.tflite
Our tflite anchors file is stored in <model_name>/model_android/anchors.json
Our tflite labels file is stored in <model_name>/model_android/labels.json

cd examples/tflite_interpreter/basic/
python python-tflite.py --MODEL_DIR /path/to/<model_name>/model_android

This script calls the tflite model interpreter for inference on all .jpg files inside the directory PATH_TO_TEST_IMAGES_DIR.

Similary the output .jpg files are storesd in PATH_TO_OUTPUT_DIR.

We can also specify the minimum confidence (score) for a given detection box to be displayed with MINIMUM_CONFIDENCE.

Finally:

python python-tflite.py \
--MODEL_DIR /path/to/<model_name>/model_android \
--PATH_TO_TEST_IMAGES_DIR /path/to/test/images \
--PATH_TO_OUTPUT_DIR /path/to/output/images \
--MINIMUM_CONFIDENCE 0.01

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites and installing

Install the required packages in requirement.txt

Creating a new virtual environement is recommended.

conda create -n object_detection python=3.7
conda activate object_detection

Git clone the repo and change directory into it. Then pip install the packages in requirement.txt.

cd directory/you/want/to/clone/into
git clone https://github.com/cloud-annotations/object-detection-python.git
cd object-detection-python
pip install -r requirement.txt

Test if everything is working

I have supplied a test model and some test images. This should output the images with detection boxes and labels in jpg format in 'examples/tflite_interpreter/basic/model/output'

cd examples/tflite_interpreter/basic/
python python-tflite.py

License

This project is licensed under the MIT License - see the LICENSE.md file for details.