Raspberry pi TensorFlow-lite Object detection

Raspberry pi TensorFlow-lite Object detection

How to use TensorFlow Lite object detection models on the Raspberry Pi.



TensorFlow Lite is an optimized framework for deploying lightweight deep learning models on resource-constrained edge devices. TensorFlow Lite models have faster inference time and require less processing power, so they can be used to obtain faster performance in realtime applications.


  

System Requirements

Hardware

·     Raspberry Pi Board (4B , 3B+)

·     IP Camera, USB Camera or Pi Camera

·     SD Card 32GB

·     5V DC. 2A Power Supply

Software

·     OS Raspbien 10 ( buster )

·     Python 3.7.2

·     Tensorflow lite 1.14.0

·     OpenCV 4.0.0

Machine Learning Model : MobileNet SSD V2 Quantised

 

Source Code
https://github.com/EdjeElectronics/TensorFlow-Lite-Object-Detection-on-Android-and-Raspberry-Pi


Reference

Install Tensorflow on Raspberry pi
https://www.tensorflow.org/install/source_rpi

Tensorflow Model Zoo
https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md

How to Run TensorFlow Lite Object Detection

https://github.com/EdjeElectronics/TensorFlow-Lite-Object-Detection-on-Android-and-Raspberry-Pi/blob/master/Raspberry_Pi_Guide.md

People tracking and counting with Raspberry pi 4 and Intel neural compute stick

People tracking and counting with Raspberry pi 4 and Intel neural compute stick

or Pedestrian Tracker

Hardware

▪ Raspberry pi 4B ( 1GB )
▪ Intel Neural Compute Stick 2
▪ 32GB SD card

Software

▪ Raspbien 10 ( buster )

▪ OpenVINO toolkit 2019 R1

▪ OpenCV 4.0.0

▪ Code C/C++

▪ Model : Person-detection-retail-0013

▪ Model : Person-reidentification-0031

▪ Video MP4 768 x 432 12 fps

 

What is Intel Neural Compute Stick and OpenVINO?

https://raspberrypi4u.blogspot.com/2019/04/raspberry-pi-openvino-intel-movidius.html

Pedestrian Tracker C++ Demo

This demo showcases Pedestrian Tracking scenario: it reads frames from an input video sequence, detects pedestrians in the frames, and builds trajectories of movement of the pedestrians in a frame-by-frame manner. You can use a set of the following pre-trained models with the demo:

• person-detection-retail-0013, which is the primary detection network for finding pedestrians
• person-reidentification-retail-0031, which is the network that is executed on top of the results from inference of the first network and makes reidentification of the pedestrians

Source Code

https://docs.openvinotoolkit.org/latest/_demos_pedestrian_tracker_demo_README.html

Reference

Install the Intel® Distribution of OpenVINO™ Toolkit for Raspbian* OS

https://software.intel.com/en-us/articles/OpenVINO-Install-RaspberryPI

Pretrained Models

https://software.intel.com/en-us/openvino-toolkit/documentation/pretrained-models

Inference Engine Samples 

http://docs.openvinotoolkit.org/latest/_docs_IE_DG_Samples_Overview.html

OpenVINO, OpenCV, and Movidius NCS on the Raspberry Pi

https://www.pyimagesearch.com/2019/04/08/openvino-opencv-and-movidius-ncs-on-the-raspberry-pi/

Car Detection with Raspberry pi 4 + Intel Neural Compute Stick

Car Detection with Raspberry pi 4 + Intel Neural Compute Stick

System Requirements

Hardware:

• Raspberry Pi 4B board or 3B+ 
• 32GB microSD card

• One of Intel® Movidius™ Visual Processing Units (VPU):

Intel® Movidius™ Neural Compute Stick or Intel® Neural Compute Stick 2

Operating Systems:

• Raspbian 10 ( Buster )
• OpenVINO for Raspberry pi ( 2019.1.094 )
• OpenCV 4.0.0
• Python 3.7.3

Machine Learning

Model : SSD MobileNet V2

Video Test

MP4 960x540 Resolution

Source Code in Python
https://docs.openvinotoolkit.org/latest/_inference_engine_ie_bridges_python_sample_object_detection_demo_ssd_async_README.html

Run Code

Test 1 Raspberry pi 4B  NCS2

Test 2 Raspberry pi 4B  NCS1

Test 3 Raspberry pi 3B+  NCS2

Test 4 Raspberry pi 3B+  NCS1 

Test Result

Raspberry Pi 4 vs Raspberry pi 3B+

Reference

Install the Intel® Distribution of OpenVINO™ Toolkit for Raspbian* OS

https://software.intel.com/en-us/articles/OpenVINO-Install-RaspberryPI

Pretrained Models

https://software.intel.com/en-us/openvino-toolkit/documentation/pretrained-models

Inference Engine Samples 

http://docs.openvinotoolkit.org/latest/_docs_IE_DG_Samples_Overview.html

OpenVINO, OpenCV, and Movidius NCS on the Raspberry Pi

https://www.pyimagesearch.com/2019/04/08/openvino-opencv-and-movidius-ncs-on-the-raspberry-pi/

Raspberry pi 4 TensorFlow Face Recognition

Raspberry pi 4 TensorFlow Face Recognition

Hardware
Raspberry pi 4B - 1GB , Raspberry pi 3B+
SD card 32 GB.

Software
Raspbien 10 ( buster )
TensorFlow 1.13.1
OpenCV 4.0.0
Python 3.7.3

Machine Learning 
Model : Facenet Inception Resnet V1

Source Code
FaceRec.  A simple working facial recognition program.
https://github.com/vudung45/FaceRec 

Recognition Dataset

Run Code on Raspberry pi 4 , 3B+ in Image

Recognise with no dataset

Run Code on Macbook Pro 2.3 GHz Intel Core i7 ( 2012 )

Reference

FaceRec.  A simple working facial recognition program.
https://github.com/vudung45/FaceRec

Pretrained models from: https://github.com/davidsandberg/facenet

Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks
https://kpzhang93.github.io/MTCNN_face_detection_alignment/

Face recognition with OpenCV, Python, and deep learning

https://www.pyimagesearch.com/2018/06/18/face-recognition-with-opencv-python-and-deep-learning/