Step by Step on how to create an vision recognition model using LOBE.ai, export the model and run the model in an Azure Function

Overview

Vision recognition using LOBE AI and Azure Functions

License: MIT Twitter: elbruno GitHub: elbruno

During the last couple of months, I’ve having fun with my new friends at home: 🐿️ 🐿️ 🐿️ . These little ones, are extremelly funny, and they literally don’t care about the cold 🥶 ❄️ ☃️ .

So, I decided to help them and build an Automatic Feeder using Azure IoT, a Wio Terminal and maybe some more devices. You can check the Azure IoT project here Azure IoT - Squirrel Feeder.

Once the feeder was ready, I decided to add a new feature to the scenario, detecting when a squirrel 🐿️ is nearby the feeder. In this repository I'll share:

  • How to create an image recognition model using LOBE.
  • How to export the model to a TensorFlow image format.
  • How to run the model in an Azure Function.

LOBE AI

LOBE is a free, easy-to-use Microsoft desktop application that allows you to build, manage, and use custom machine learning models. With Lobe, you can create an image classification model to categorize images into labels that represent their content.

Here's a summary of how to prepare a model in Lobe:

  • Import and label images.
  • Train your model.
  • Evaluate training results.
  • Play with your model to experiment with different scenarios.
  • Export and use your model in an app.

The Overview of image classification model by Lobe section contains step-by-step instructions that let you make calls to the service and get results in a short period of time.

You can use the images in the "LOBE/Train/" directory in this repository to train your model.

Here is the model performing live recognition in action:

Exporting the model to TensorFlow

Once the project was trained, you can export it to several formats. We will use a TensorFlow format for the Azure Function.

The exported model has several files. The following list shows the files that we use in our Azure Function:

  • labels.txt: The labels that the model recognizes
  • saved_model.pb: The model definition
  • signature.json: The model signature
  • example/tf_example.py.py: sample python code that uses the exported model.

You can check the exported model in the "Lobe/ExportedModel" directory in this repository.

Azure Function

Time to code! Let's create a new Azure Function Using Visual Studio Code and the Azure Functions for Visual Studio Code extension.

Changes to __ init __.py

The following code is the final code for the __ init __.py file in the Azure Function.

A couple of notes:

  • The function will receive a POST request with the file bytes in the body.
  • In order to use the tf_model_helper file, we must import the tf_model_helper.py function from the tf_model_helper.py file using ".tf_model_helper"
  • ASSETS_PATH and TF_MODEL are the variables that we will use to access the exported model. We will use os.path to resolve the current path to the exported model.
  • The result of the function will be a JSON string with the prediction. Jsonify will convert the TF_Model() image prediction to a JSON string.
import logging
import azure.functions as func

# Imports for image procesing
import io
import os
from PIL import Image
from flask import Flask, jsonify

# Imports for prediction
from .tf_model_helper import TFModel

def main(req: func.HttpRequest) -> func.HttpResponse:
    logging.info('Python HTTP trigger function processed a request.')

    results = "{}"
    try:
        # get and load image from POST
        image_bytes = req.get_body()    
        image = Image.open(io.BytesIO(image_bytes))

        # Load and intialize the model and the app context
        app = Flask(__name__)  

        # load LOBE Model using the current directory
        scriptpath = os.path.abspath(__file__)
        scriptdir  = os.path.dirname(scriptpath)
        ASSETS_PATH = os.path.join(scriptdir, "model")
        TF_MODEL = TFModel(ASSETS_PATH)

        with app.app_context():        
            # prefict image and process results in json string format
            results = TF_MODEL.predict(image)            
            jsonresult = jsonify(results)
            jsonStr = jsonresult.get_data(as_text=True)
            results = jsonStr

    except Exception as e:
        logging.info(f'exception: {e}')
        pass 

    # return results
    logging.info('Image processed. Results: ' + results)
    return func.HttpResponse(results, status_code=200)

Changes to requirements.txt

The requirements.txt file will define the necessary libraries for the Azure Function. We will use the following libraries:

# DO NOT include azure-functions-worker in this file
# The Python Worker is managed by Azure Functions platform
# Manually managing azure-functions-worker may cause unexpected issues

azure-functions
requests
Pillow
numpy
flask
tensorflow
opencv-python

Sample Code

You can view a sample function completed code in the "AzureFunction/LobeSquirrelDetectorFunction/" directory in this repository.

Testing the sample

Once our code is complete we can test the sample in local mode or in Azure Functions, after we deploy the Function. In both scenarios we can use any tool or language that can perform HTTP POST requests to the Azure Function to test our function.

Test using Curl

Curl is a command line tool that allows you to send HTTP requests to a server. It is a very simple tool that can be used to send HTTP requests to a server. We can test the local function using curl with the following command:

❯ curl http://localhost:7071/api/LobeSquirrelDetectorFunction -Method 'Post' -InFile 01.jpg

Test using Postman

Postman is a great tool to test our function. You can use it to test the function in local mode and also to test the function once it has been deployed to Azure Functions. You can download Postman here.

In order to test our function we need to know the function url. In Visual Studio Code, we can get the url by clicking on the Functions section in the Azure Extension. Then we can right click on the function and select "Copy Function URL".

Now we can go to Postman and create a new POST request using our function url. We can also add the image we want to test. Here is a live demo, with the function running locally, in Debug mode in Visual Studio Code:

We are now ready to test our function in Azure Functions. To do so we need to deploy the function to Azure Functions. And use the new Azure Function url with the same test steps.

Additional Resources

You can check a session recording about this topic in English and Spanish.

These links will help to understand specific implementations of the sample code:

In my personal blog "ElBruno.com", I wrote about several scenarios on how to work and code with LOBE.

Author

👤 Bruno Capuano

🤝 Contributing

Contributions, issues and feature requests are welcome!

Feel free to check issues page.

Show your support

Give a ⭐️ if this project helped you!

📝 License

Copyright © 2021 Bruno Capuano.

This project is MIT licensed.


Owner
El Bruno
Sr Cloud Advocate @Microsoft, former Microsoft MVP (14 years!), lazy runner, lazy podcaster, technology enthusiast
El Bruno
[ICCV 2021 Oral] NerfingMVS: Guided Optimization of Neural Radiance Fields for Indoor Multi-view Stereo

NerfingMVS Project Page | Paper | Video | Data NerfingMVS: Guided Optimization of Neural Radiance Fields for Indoor Multi-view Stereo Yi Wei, Shaohui

Yi Wei 369 Dec 24, 2022
Diverse Image Generation via Self-Conditioned GANs

Diverse Image Generation via Self-Conditioned GANs Project | Paper Diverse Image Generation via Self-Conditioned GANs Steven Liu, Tongzhou Wang, David

Steven Liu 147 Dec 03, 2022
Finite-temperature variational Monte Carlo calculation of uniform electron gas using neural canonical transformation.

CoulombGas This code implements the neural canonical transformation approach to the thermodynamic properties of uniform electron gas. Building on JAX,

FermiFlow 9 Mar 03, 2022
IA for recognising Traffic Signs using Keras [Tensorflow]

Traffic Signs Recognition ⚠️ 🚦 Fundamentals of Intelligent Systems Introduction 📄 Development of a neural network capable of recognizing nine differ

Sebastián Fernández García 2 Dec 19, 2022
Perfect implement. Model shared. x0.5 (Top1:60.646) and 1.0x (Top1:69.402).

Shufflenet-v2-Pytorch Introduction This is a Pytorch implementation of faceplusplus's ShuffleNet-v2. For details, please read the following papers:

423 Dec 07, 2022
Unsupervised Foreground Extraction via Deep Region Competition

Unsupervised Foreground Extraction via Deep Region Competition [Paper] [Code] The official code repository for NeurIPS 2021 paper "Unsupervised Foregr

28 Nov 06, 2022
Metadata-Extractor - Metadata Extractor Script can be used to read in exif metadata

Metadata Extractor The exifextract script can be used to read in exif metadata f

1 Feb 16, 2022
The source code for CATSETMAT: Cross Attention for Set Matching in Bipartite Hypergraphs

catsetmat The source code for CATSETMAT: Cross Attention for Set Matching in Bipartite Hypergraphs To be able to run it, add catsetmat to PYTHONPATH H

2 Dec 19, 2022
Circuit Training: An open-source framework for generating chip floor plans with distributed deep reinforcement learning

Circuit Training: An open-source framework for generating chip floor plans with distributed deep reinforcement learning. Circuit Training is an open-s

Google Research 479 Dec 25, 2022
UniMoCo: Unsupervised, Semi-Supervised and Full-Supervised Visual Representation Learning

UniMoCo: Unsupervised, Semi-Supervised and Full-Supervised Visual Representation Learning This is the official PyTorch implementation for UniMoCo pape

dddzg 49 Jan 02, 2023
Deduplicating Training Data Makes Language Models Better

Deduplicating Training Data Makes Language Models Better This repository contains code to deduplicate language model datasets as descrbed in the paper

Google Research 431 Dec 27, 2022
Code repo for "Cross-Scale Internal Graph Neural Network for Image Super-Resolution" (NeurIPS'20)

IGNN Code repo for "Cross-Scale Internal Graph Neural Network for Image Super-Resolution" [paper] [supp] Prepare datasets 1 Download training dataset

Shangchen Zhou 278 Jan 03, 2023
A fast poisson image editing implementation that can utilize multi-core CPU or GPU to handle a high-resolution image input.

Poisson Image Editing - A Parallel Implementation Jiayi Weng (jiayiwen), Zixu Chen (zixuc) Poisson Image Editing is a technique that can fuse two imag

Jiayi Weng 110 Dec 27, 2022
Rule-based Customer Segmentation

Rule-based Customer Segmentation Business Problem A game company wants to create level-based new customer definitions (personas) by using some feature

Cem Çaluk 2 Jan 03, 2022
Convert game ISO and archives to CD CHD for emulation on Linux.

tochd Convert game ISO and archives to CD CHD for emulation. Author: Tuncay D. Source: https://github.com/thingsiplay/tochd Releases: https://github.c

Tuncay 20 Jan 02, 2023
PromptDet: Expand Your Detector Vocabulary with Uncurated Images

PromptDet: Expand Your Detector Vocabulary with Uncurated Images Paper Website Introduction The goal of this work is to establish a scalable pipeline

103 Dec 20, 2022
Tzer: TVM Implementation of "Coverage-Guided Tensor Compiler Fuzzing with Joint IR-Pass Mutation (OOPSLA'22)“.

Artifact • Reproduce Bugs • Quick Start • Installation • Extend Tzer Coverage-Guided Tensor Compiler Fuzzing with Joint IR-Pass Mutation This is the s

12 Dec 29, 2022
Code of our paper "Contrastive Object-level Pre-training with Spatial Noise Curriculum Learning"

CCOP Code of our paper Contrastive Object-level Pre-training with Spatial Noise Curriculum Learning Requirement Install OpenSelfSup Install Detectron2

Chenhongyi Yang 21 Dec 13, 2022
Code Repo for the ACL21 paper "Common Sense Beyond English: Evaluating and Improving Multilingual LMs for Commonsense Reasoning"

Common Sense Beyond English: Evaluating and Improving Multilingual LMs for Commonsense Reasoning This is the Github repository of our paper, "Common S

INK Lab @ USC 19 Nov 30, 2022