MatchingService

• Service based AI
• The service designed to serve Indoor Localization App
• It has a couple of APIs, Enrich, Find and Train.
  • Enrich API, to enable the APP enrich database with images and metadata on locations, by uploading video that describes the location.
  • In Enrich API, there is an ability to label existing data, by uploading image to label specific location in database.
  • Finds API, is an API that uses ML trained model, to find the location of uploaded image.
  • Train API, is an API to train the ML model upon the data existing in database,

Indoor Localization Backend

To run app, run the command in terminal/cmd:

python app.py

Api Calls Examples:

Enrichment API

Video:

curl --location 'http://{host}:{port}/api/enrich/video' \
--form 'video=@"20230719_094627.mp4"' \
--form 'name="The Phoenicians "' \
--form 'description="Archaeology"'

Label:

curl --location 'http://{host}:{port}/api/enrich/label' \
--form 'image=@"20230719_094627/image1.jpg"' \
--form 'class="648f39d5e60ce40af4d51d1c"'

Find API

Train Model API

curl --location 'http://{host}:{port}/api/train'

HealthCheck API

curl --location 'http://{host}:{port}/api/healthcheck'

Service Architecture

TODO:

• Build training data in this: get top 3 results, get percent of results above threshold (mini research needed to decide about it) and add column that will contain the location zone. example data frame should look like: [top1_clip, top1_resnet, top2_clip, top2_resnet, top3_clip, top3_resnet, percent, zone]
• Capture a video from museum, then split it to frames, these frames compare them to the representative images that image_selector chose, then check if frames that captured should be represented by or not.
• In thesis proposal, update the Research question and goals by deleting the repetitive text, and replace it by talking on challenges.
• Need also to add primer results, to talk about results we have till now.