Tensorflow: Detector 2D (Python API)

• Introduction
• Trained Models
• Running Software
• Requirements
• Status
• Related projects

Introduction

tensorflowLiteDetection2D module use tensorflow lite Python API. The module detects using pre-trained models like VGG16, Inception, ResNet, MobileNet and other pre-trained models. Also use YARP to send video source pre and post-procesed. Also admits YARP source video like input. The module has been tested with ssd_mobilenet_v1_1.0_coco. This module also publish detection results in YARP port with object coordinates.

Figure 1. Detection example

Documentation available on docs.

Trained Models

tensorflowLiteDetection2D requires a model to detect. Models should be located in models dir.

Running Software

1. Create or configure YARP Server.

yarp server

yarp conf pc-ip 10000

2. Execute programs/tensorflowLiteDetection2D.py the detector.

python3 tensorflowLiteDetection2D.py

3. Connect video source to tensorflowLiteDetection2D.

yarp connect /videoSource /tensorflowLiteDetection2D/img:i

NOTE:

• Video results are published on /tensorflowLiteDetection2D/img:o
• Data results are published on /tensorflowLiteDetection2D/data:o
• Coordinate results are published on /tensorflowLiteDetection2D/coord:o

Requirements

tensorflowLiteDetection2D requires:

• Install OpenCV

pip3 install opencv-python==3.4.6.27

• Install YARP 2.3.XX+ with Python 3 bindings
• Install pip3
• Install tensorflow:

pip3 install tensorflow

Tested on: ubuntu 14.04, ubuntu 16.04, ubuntu 18.04, lubuntu 18.04 and raspbian.

Status

Related projects

• ageitgey: face_recognition project
• davidvelascogarcia: Tensorflow: Detector 2D (C++ API)
• EdjeElectronics: TensorFlow-Lite-Object-Detection-on-Android-and-Raspberry-Pi