Please enable JavaScript.

Coggle requires JavaScript to display documents.

CPAW - Coggle Diagram

- - - - construct fetchHandler with arguments:
        UploadDataProvider instance,
        GeneralClassifier instance,
      - API: getNDataURLs
        return pair where first is an array of data urls for images
        and second is an array of imkey,obkey pairs. take in the fetchType whether random, positive, or negative
        
        Random
        
        get N random obkey,imkey pairs from UploadDataProvider instance API call: getRandomNPairs()
        
        input the random obkey,imkey pairs into function calls to
        pairToDataURL() to get data urls and return array of them
        
        Positive/negative
        
        get N random obkey, imkey pairs
        
        get N random obkey,imkey pairs from UploadDataProvider instance API call: getRandomNPairs()
        
        get the corresponding rows of object data and pass them into the classifier to get their predicted labels
        
        use UploadDataProvider instance's pairToObjectDataRow() method calls to make an array of data objects corresponding to the array of obkey, imkeys
        
        get the column names that we care about for machine learning with function call to UploadDataProvider instance with method getFeaturesToUse()
        
        Pass data objects and column names into GeneralClassifier instance's predict function and get an array of 0s and 1s meaning whether each is negative or positive respectively
        
        add the imkey, obkeys corresponding to positive/negative labels to the return array
        
        loop until return array is N or we reach the max number of iterations (let's make it 10 for now)
        
        after the loop is done,
        input the random obkey,imkey pairs into function calls to
        pairToDataURL() to get data urls and return array of them
    - - construct with a list of dataurls and a list of pairs of the same length where each pair has one image key and one object key. Instance variables need to include list of dataurls, list of pairs, and length of the list
      - Each image object starts out with "unclassified" field , when drag and dropped give new tag whether now in positive or negative grid
      - API: get imkeys, obkey pairs for: unclassified field, negative field, and positive field in image data state
      - API: get data url by index
      - API: set grid placement property by index and gridType
      - API: get number of images
      - API: set internal list of imkey, obkey pairs with input of list of indices and the gridType
    - - Download Result
        
        Download classifier.json spec
        
        userDownloadJSON function
        
        Download newtrainingset.txt
        
        userDownloadText function
      - Construct Classifier
        
        Fit to Dataset
        
        convert dataset to tensors
        
        start training
        
        input dataset, labels, column_names
        
        save training set for later download
        
        LogisticRegression
      - Predict Dataset
        
        take in subarray of object data, and column names array of strings
        
        output array of labels (0 and 1 for negative and positive)
      - render confusion matrix
    - - Make UploadDataProvider class
        
        API: get random N rows of object_data
        
        API: get random row of object_data
        
        Initialize on upload file
        
        API: get initial training data set
        
        API: getRandomNPairs() get random N (imkey,obkey) pairs
        
        API: pairToDataURL() return data URL for merged 3 channel image by imkey and obkey
        
        construct 3 channel image objects by imkey
        
        store in canvas, slicing by obkey derived nuclei location
        
        convert canvas data url and remove canvas
        
        API: pairToObjectDataRow(); get row of object_data by imkey and obkey
        
        API: getFeaturesToUse() get features to use (column names)
    - - Basic Easy Sequence
        
        DataUploadedStandby State
        
        Allow Fetch Button Only
        
        Construct UploadDataProvider
        
        construct one instance of GeneralClassifier with initial training set from UploadDataProvider
        
        call CoreIterationManager.onUploadButton() , fileListObject argument
        
        Fetching Image State
        
        Fetching Random State
        
        Construct FetchHandler with GeneralClassifier and UploadDataProver instances as argument
        
        get random N data dataurls from FetchHandler
        and pass into construction of Image Grid Manager
        
        Allow No Interaction
        
        Fetching Positive/Negative State
        
        Allow No interaction
        
        Cons
        
        call CoreIterationManager.onFetchButton(),
        
        Training Model State
        
        Allow No Interaction
        
        set grid types of list of indices in
        ImageGridManager
        
        get list of imkey obkey pairs for positive and negative bins
        
        get subarray of object_data which corresponds to those imkey obkeys and construct array of 0s and 1s where 0s correspond to negative binned and 1s to positive
        
        Construct GeneralClassifier with input of this subarray and label array
        
        DragandDrop State
        
        RandomFetchComplete State
        
        Allow Dragging between
        unclassified grid and both
        the positive and negative state
        
        PositiveFetchComplete State
        
        Allow Dragging between
        unclassified grid and
        negative grid
        
        NegativeFetchComplete State
        
        Allow Dragging between
        unclassified grid and
        positive grid
        
        Initial State
        
        Allow Upload Button Only
        
        Construct This class on app load