Please enable JavaScript.

Coggle requires JavaScript to display documents.

Activities (WP3 & WP5) - Coggle Diagram

- - - - Legume class
        
        Secondary crops:
        groundnut
        soybean
        chickpea
        
        Primary Crops:
        Common Bean
        Cowpea
        
        SOPs & traits
        
        Stand Count
        
        Flower Count
        
        Pod Count
        
        Disease (PHI)
      - Grass
        
        Primary Crops:
        Sorghum
        
        Secondary crops:
        Maize
        Millet
        
        SOPs & traits
        
        Panicle formation
        
        Stand Count
    - - which image types is it worth doing for?
      - distance tracking mechanisms (using wheel rotations to automate image capture so each image is of a given distance and minimizes overlap?)
      - easier plot selection instead of taking the phones off the holders
        
        have a third screen to show and control what's happening on the other 2 screens
  - - - time taken
        
        average across development
        
        accounts for variance between people, years of experience, improvement from start of season to end of season, bias from time of day/tiredness
        
        eliminates issues with
        
        at one point for given distance or number of plants
        
        much quicker
        
        time taken per plot
      - number of people per SOP
    - - correlations
        
        manual phenotyping between raters
        
        imagery based phenotyping correlations to GT
- - - - drop image types that are not working back into the SOP development process or even completely if necessary
    - - To Test
        
        Model performance across crop types (legumes, grass)
        
        Model perfomance across traits
        
        Model performance with different approaches [model architectures (CNN, foundational. ag-specific)]
        
        Zero-shot, few-shot, transfer learning (can PC models trabsfer to SC)
        
        Model performance across image collection method and image type (Bruno vs manual, panorama vs others)
  - - - Existing SOPs
        
        test good SOPs on low cost phones
        
        Test PC SOPs on SC to see transferability
      - Testing strategies:
        
        visual assessment of trait using images
        
        quality check
        
        efficiency checks (time, number of people)
        
        easy to learn?
        
        Automated image assessment approaches
      - NEW SOPs
        
        Trait to image and image to trait brainstorming & testing (quick iterations of testing)
        
        Employ testing strategies