miguelpeixe/aetrapp-ips

•Updated over 8 years ago

AETrapp Image Processing Service

Image

10K+

miguelpeixe/aetrapp-ips repository overview

⁠AeTrapp Image Processing Service

This project is part of AeTrapp⁠ - a collaborative effort designed for community engagement in monitoring population of Aedes aegypti mosquitoes, which is a known vector of several diseases including dengue, Zika, chikungunya and urban yellow fever. The idea behind this initiative is to provide an easy-to-use mobile app as a tool for getting instructions to create your own traps, capture samples containing eggs of Aedes, share them to the community and to acquire informations about its incidence in a region shared by other users.

Original sample (left), sample after recognition (right)

⁠Original sample (left) and after recognition process (right)

Check our contributing guide⁠ and feel free to open issues.

⁠How it works?

This repository contains source code for image processing tasks applied to automatic detection and counting of eggs of Aedes aegypti. The activity chart below shows a summary of its operation.

Dalek Sequence

Pre-Processing routines are responsible for checking the quality of an input image such its positioning, blur rate, shadow overlay, etc.
Segmentation routines driscriminate relevant objects (eggs or cluster of eggs) from the background.
Eggs and clusters are currently recognized by their size, shape and color.

⁠Download the Image Database

Link for download: https://drive.google.com/open?id=15YtFv4kor01GsTeiZGsP3Wi8au78udQm⁠

⁠More from AeTrapp project

Mobile: https://github.com/aetrapp/mobile⁠.
WebApp: https://github.com/aetrapp/webapp⁠.
API: https://github.com/aetrapp/api⁠.

⁠Installing

Node.js 8+ and Docker are required. Clone this repository locally and download the model:

    git clone <repository_url>
    cd <repository_path>
    ./download-model.sh

Build and start the server with Docker:

docker-compose up

Then, access the dashboard at http://localhost:3131/agenda⁠. To create your first job, make a POST request to /agenda/api/jobs/create like in the following example. You can change "imageUrl" value to any image available:

curl -H "Content-Type: application/json" -X POST -d \
    '{
      "jobName": "process image",
      "jobSchedule": "now",
      "jobData": {
         "image": {
             "url": "https://github.com/aetrapp/image-processing-service/raw/master/samples/06.4SEM.CENC.INTRA.SONY.jpg"
         }
      }
     }' http://localhost:3131/agenda/api/jobs/create

If you need shell access to the server, get your container name with docker ps and run:

docker exec -i -t <imageprocessingservice_container_name> /bin/bash

⁠Deploying new version to production

AeTrapp's main API server checks every five minutes the current version of the Image Processing Server, and will perform re-analysis of samples if a new version is found. To start this process:

Change algorithm's code and commit;
Change ipsVersion value in package.json and commit:

{
  "name": "image-processing-service",
  "version": "0.0.1",
  "ipsVersion": "1.4.0", <- CHANGE THIS LINE
  "description": "Aetrapp Image Processing Service",
...