Create a Skin Condition classification task

.gitignore

@@ -1,7 +1,8 @@
 # dependencies
 /node_modules/
-# disco models
-**/models
+# model.json files
+server/models
+cli/models
 # built
 dist/
 

discojs/src/dataset/dataset_builder.ts

@@ -13,16 +13,11 @@ export class DatasetBuilder<Source> {
   /**
    * The buffer of unlabelled file sources.
    */
-  private _sources: Source[]
+  private _unlabeledSources: Source[]
   /**
    * The buffer of labelled file sources.
    */
-  private labelledSources: Map<string, Source[]>
-  /**
-   * Whether a dataset was already produced.
-   */
-  // TODO useless, responsibility on callers
-  private _built: boolean
+  private _labeledSources: Map<string, Source[]>
 
   constructor (
     /**
@@ -34,9 +29,9 @@ export class DatasetBuilder<Source> {
      */
     public readonly task: Task
   ) {
-    this._sources = []
-    this.labelledSources = Map()
-    this._built = false
+    this._unlabeledSources = []
+    // Map from label to sources
+    this._labeledSources = Map()
   }
 
   /**
@@ -46,17 +41,14 @@ export class DatasetBuilder<Source> {
    * @param label The file sources label
    */
   addFiles (sources: Source[], label?: string): void {
-    if (this.built) {
-      this.resetBuiltState()
-    }
     if (label === undefined) {
-      this._sources = this._sources.concat(sources)
+      this._unlabeledSources = this._unlabeledSources.concat(sources)
     } else {
-      const currentSources = this.labelledSources.get(label)
+      const currentSources = this._labeledSources.get(label)
       if (currentSources === undefined) {
-        this.labelledSources = this.labelledSources.set(label, sources)
+        this._labeledSources = this._labeledSources.set(label, sources)
       } else {
-        this.labelledSources = this.labelledSources.set(label, currentSources.concat(sources))
+        this._labeledSources = this._labeledSources.set(label, currentSources.concat(sources))
       }
     }
   }
@@ -67,28 +59,19 @@ export class DatasetBuilder<Source> {
    * @param label The file sources label
    */
   clearFiles (label?: string): void {
-    if (this.built) {
-      this.resetBuiltState()
-    }
     if (label === undefined) {
-      this._sources = []
+      this._unlabeledSources = []
     } else {
-      this.labelledSources = this.labelledSources.delete(label)
+      this._labeledSources = this._labeledSources.delete(label)
     }
   }
 
-  // If files are added or removed, then this should be called since the latest
-  // version of the dataset_builder has not yet been built.
-  private resetBuiltState (): void {
-    this._built = false
-  }
-
   private getLabels (): string[] {
     // We need to duplicate the labels as we need one for each source.
     // Say for label A we have sources [img1, img2, img3], then we
     // need labels [A, A, A].
     let labels: string[][] = []
-    this.labelledSources.forEach((sources, label) => {
+    this._labeledSources.forEach((sources, label) => {
       const sourcesLabels = Array.from({ length: sources.length }, (_) => label)
       labels = labels.concat(sourcesLabels)
     })
@@ -97,56 +80,46 @@ export class DatasetBuilder<Source> {
 
   async build (config?: DataConfig): Promise<DataSplit> {
     // Require that at least one source collection is non-empty, but not both
-    if ((this._sources.length > 0) === (this.labelledSources.size > 0)) {
-      throw new Error('Please provide dataset input files') // This error message is parsed in DatasetInput.vue
+    if (this._unlabeledSources.length + this._labeledSources.size === 0) {
+      throw new Error('No input files connected') // This error message is parsed in Trainer.vue
     }
-
     let dataTuple: DataSplit
-    if (this._sources.length > 0) {
+    if (this._unlabeledSources.length > 0) {
       let defaultConfig: DataConfig = {}
 
       if (config?.inference === true) {
         // Inferring model, no labels needed
         defaultConfig = {
           features: this.task.trainingInformation.inputColumns,
-          shuffle: false
+          shuffle: true
         }
       } else {
         // Labels are contained in the given sources
         defaultConfig = {
           features: this.task.trainingInformation.inputColumns,
           labels: this.task.trainingInformation.outputColumns,
-          shuffle: false
+          shuffle: true
         }
       }
 
-      dataTuple = await this.dataLoader.loadAll(this._sources, { ...defaultConfig, ...config })
+      dataTuple = await this.dataLoader.loadAll(this._unlabeledSources, { ...defaultConfig, ...config })
     } else {
       // Labels are inferred from the file selection boxes
       const defaultConfig = {
         labels: this.getLabels(),
-        shuffle: false
+        shuffle: true
       }
-      const sources = this.labelledSources.valueSeq().toArray().flat()
+      const sources = this._labeledSources.valueSeq().toArray().flat()
       dataTuple = await this.dataLoader.loadAll(sources, { ...defaultConfig, ...config })
     }
-    // TODO @s314cy: Support .csv labels for image datasets (supervised training or testing)
-    this._built = true
     return dataTuple
   }
 
-  /**
-   * Whether the dataset builder has already been consumed to produce a dataset.
-   */
-  get built (): boolean {
-    return this._built
-  }
-
   get size (): number {
-    return Math.max(this._sources.length, this.labelledSources.size)
+    return Math.max(this._unlabeledSources.length, this._labeledSources.size)
   }
 
   get sources (): Source[] {
-    return this._sources.length > 0 ? this._sources : this.labelledSources.valueSeq().toArray().flat()
+    return this._unlabeledSources.length > 0 ? this._unlabeledSources : this._labeledSources.valueSeq().toArray().flat()
   }
 }

discojs/src/default_tasks/cifar10/index.ts → discojs/src/default_tasks/cifar10.ts

@@ -1,9 +1,9 @@
 import * as tf from '@tensorflow/tfjs'
 
-import type { Model, Task, TaskProvider } from '../../index.js'
-import { data, models } from '../../index.js'
+import type { Model, Task, TaskProvider } from '../index.js'
+import { data, models } from '../index.js'
 
-import baseModel from './model.js'
+import baseModel from '../models/mobileNet_v1_025_224.js'
 
 export const cifar10: TaskProvider = {
   getTask (): Task {

discojs/src/default_tasks/index.ts

@@ -1,6 +1,7 @@
-export { cifar10 } from './cifar10/index.js'
+export { cifar10 } from './cifar10.js'
 export { lusCovid } from './lus_covid.js'
+export { skinCondition } from './skin_condition.js'
 export { mnist } from './mnist.js'
-export { simpleFace } from './simple_face/index.js'
+export { simpleFace } from './simple_face.js'
 export { titanic } from './titanic.js'
 export { wikitext } from './wikitext.js'

discojs/src/default_tasks/lus_covid.ts

@@ -11,7 +11,7 @@ export const lusCovid: TaskProvider = {
         taskTitle: 'COVID Lung Ultrasound',
         summary: {
           preview: 'Do you have a data of lung ultrasound images on patients <b>suspected of Lower Respiratory Tract infection (LRTI) during the COVID pandemic</b>? <br> Learn how to discriminate between COVID positive and negative patients by joining this task.',
-          overview: "Don’t have a dataset of your own? Download a sample of a few cases <a class='underline' href='https://drive.switch.ch/index.php/s/zM5ZrUWK3taaIly' target='_blank'>here</a>."
+          overview: "Don't have a dataset of your own? Download a sample of a few cases <a class='underline' href='https://drive.switch.ch/index.php/s/zM5ZrUWK3taaIly' target='_blank'>here</a>."
         },
         model: "We use a simplified* version of the <b>DeepChest model</b>: A deep learning model developed in our lab (<a class='underline' href='https://www.epfl.ch/labs/mlo/igh-intelligent-global-health/'>intelligent Global Health</a>.). On a cohort of 400 Swiss patients suspected of LRTI, the model obtained over 90% area under the ROC curve for this task. <br><br>*Simplified to ensure smooth running on your browser, the performance is minimally affected. Details of the adaptations are below <br>- <b>Removed</b>: positional embedding (i.e. we don’t take the anatomic position into consideration). Rather, the model now does mean pooling over the feature vector of the images for each patient <br>- <b>Replaced</b>: ResNet18 by Mobilenet",
         dataFormatInformation: 'This model takes as input an image dataset. It consists on a set of lung ultrasound images per patient with its corresponding label of covid positive or negative. Moreover, to identify the images per patient you have to follow the follwing naming pattern: "patientId_*.png"',

discojs/src/default_tasks/simple_face/index.ts → discojs/src/default_tasks/simple_face.ts

@@ -1,9 +1,8 @@
 import * as tf from '@tensorflow/tfjs'
 
-import type { Model, Task, TaskProvider } from '../../index.js'
-import { data, models } from '../../index.js'
-
-import baseModel from './model.js'
+import type { Model, Task, TaskProvider } from '../index.js'
+import { data, models } from '../index.js'
+import baseModel from '../models/mobileNetV2_35_alpha_2_classes.js'
 
 export const simpleFace: TaskProvider = {
   getTask (): Task {
@@ -16,7 +15,7 @@ export const simpleFace: TaskProvider = {
           overview: 'Simple face is a small subset of face_task from Kaggle'
         },
         dataFormatInformation: '',
-        dataExampleText: 'Below you find an example',
+        dataExampleText: 'Below you can find an example',
         dataExampleImage: 'https://storage.googleapis.com/deai-313515.appspot.com/example_training_data/simple_face-example.png'
       },
       trainingInformation: {

discojs/src/default_tasks/skin_condition.ts

@@ -0,0 +1,95 @@
+import * as tf from '@tensorflow/tfjs'
+
+import type { Model, Task, TaskProvider } from '../index.js'
+import { data, models } from '../index.js'
+
+const IMAGE_SIZE = 128
+const LABELS = ['Eczema', 'Allergic Contact Dermatitis', 'Urticaria']
+
+export const skinCondition: TaskProvider = {
+  getTask (): Task {
+    return {
+      id: 'skin_condition',
+      displayInformation: {
+        taskTitle: 'Skin Condition Classification',
+        summary: {
+          preview: "Identify common skin conditions from volunteer image contributions. You can find a sample dataset of 400 images <a class='underline text-primary-dark dark:text-primary-light' href='https://storage.googleapis.com/deai-313515.appspot.com/scin_sample.zip'>here</a> or see the full <a class='underline text-primary-dark dark:text-primary-light' href='https://github.com/google-research-datasets/scin/tree/main'>SCIN dataset</a>. You can find how to download and preprocess the dataset <a class='underline text-primary-dark dark:text-primary-light' href='https://github.com/epfml/disco/blob/develop/docs/examples/scin_dataset.ipynb'>in this notebook</a>.",
+          overview: "The <a class='underline text-primary-dark dark:text-primary-light' href='https://github.com/google-research-datasets/scin/tree/main'>SCIN (Skin Condition Image Network) open access dataset</a> aims to supplement publicly available dermatology datasets from health system sources with representative images from internet users. To this end, the SCIN dataset was collected from Google Search users in the United States through a voluntary, consented image donation application. The SCIN dataset is intended for health education and research, and to increase the diversity of dermatology images available for public use. The SCIN dataset contains 5,000+ volunteer contributions (10,000+ images) of common dermatology conditions. Contributions include Images, self-reported demographic, history, and symptom information, and self-reported Fitzpatrick skin type (sFST). In addition, dermatologist labels of the skin condition are provided for each contribution. You can find more information on the dataset and classification task <a class='underline text-primary-dark dark:text-primary-light' href='https://arxiv.org/abs/2402.18545'>here</a>."
+        },
+        dataFormatInformation: "There are hundreds of skin condition labels in the SCIN dataset. For the sake of simplicity, we only include the 3 most common conditions in the sample dataset: 'Eczema', 'Allergic Contact Dermatitis' and 'Urticaria'. Therefore, each image is expected to be labeled with one of these three categories.",
+        sampleDatasetLink: 'https://storage.googleapis.com/deai-313515.appspot.com/scin_sample.zip'
+      },
+      trainingInformation: {
+        modelID: 'skin-condition-model',
+        epochs: 10,
+        roundDuration: 2,
+        validationSplit: 0.3,
+        batchSize: 8,
+        preprocessingFunctions: [data.ImagePreprocessing.Resize, data.ImagePreprocessing.Normalize],
+        dataType: 'image',
+        IMAGE_H: IMAGE_SIZE,
+        IMAGE_W: IMAGE_SIZE,
+        LABEL_LIST: LABELS,
+        scheme: 'federated',
+        noiseScale: undefined,
+        clippingRadius: undefined
+      }
+    }
+  },
+
+  async getModel(): Promise<Model> {
+    const imageChannels = 3
+    const numOutputClasses = LABELS.length
+
+    const model = tf.sequential()
+
+    model.add(
+      tf.layers.conv2d({
+        inputShape: [IMAGE_SIZE, IMAGE_SIZE, imageChannels],
+        filters: 8,
+        kernelSize: 3,
+        strides: 1,
+        kernelInitializer: 'varianceScaling',
+        activation: 'relu'
+      })
+    )
+    model.add(tf.layers.maxPooling2d({ poolSize: [2, 2]}))
+    model.add(tf.layers.dropout({ rate: 0.2 }))
+
+    const convFilters = [16, 32, 64, 128]
+    for (const filters of convFilters) {
+      model.add(
+        tf.layers.conv2d({
+          filters: filters,
+          kernelSize: 3,
+          strides: 1,
+          kernelInitializer: 'varianceScaling',
+          activation: 'relu'
+        })
+      )
+
+      model.add(tf.layers.maxPooling2d({ poolSize: [2, 2]}))
+      model.add(tf.layers.dropout({ rate: 0.2 }))
+    }
+
+    model.add(tf.layers.flatten())
+    model.add(tf.layers.dense({
+      units: 64,
+      kernelInitializer: 'varianceScaling',
+      activation: 'relu',
+    }))
+
+    model.add(tf.layers.dense({
+      units: numOutputClasses,
+      kernelInitializer: 'varianceScaling',
+      activation: 'softmax'
+    }))
+
+    model.compile({
+      optimizer: tf.train.adam(),
+      loss: 'categoricalCrossentropy',
+      metrics: ['accuracy']
+    })
+    return Promise.resolve(new models.TFJS(model))
+  }
+}

discojs/src/default_tasks/simple_face/model.ts → discojs/src/models/mobileNetV2_35_alpha_2_classes.ts

@@ -1,3 +1,5 @@
+// Source: https://storage.googleapis.com/deai-313515.appspot.com/models/mobileNetV2_35_alpha_2_classes/model.json
+// This model was converted using the tensorflow.js converter
 export default {
   format: "layers-model",
   generatedBy: "keras v2.6.0",

discojs/src/default_tasks/cifar10/model.ts → discojs/src/models/mobileNet_v1_025_224.ts

@@ -1,3 +1,4 @@
+// Source: https://storage.googleapis.com/tfjs-models/tfjs/mobilenet_v1_0.25_224/model.json
 export default {
   modelTopology: {
     keras_version: "2.1.4",

discojs/src/validation/validator.ts

@@ -21,23 +21,22 @@ export class Validator {
     }
   }
 
-  private async getLabel (ys: tf.Tensor): Promise<Float32Array | Int32Array | Uint8Array> {
-    switch (ys.shape[1]) {
-      case 1:
-        return await ys.greaterEqual(tf.scalar(0.5)).data()
-      case 2:
-        return await ys.argMax(1).data()
-      default:
-        throw new Error(`unable to reduce tensor of shape: ${ys.shape.toString()}`)
+  private async getLabel(ys: tf.Tensor): Promise<Float32Array | Int32Array | Uint8Array> {
+    // Binary classification
+    if (ys.shape[1] == 1) {
+      return await ys.greaterEqual(tf.scalar(0.5)).data()
+    // Multi-class classification
+    } else {
+      return await ys.argMax(-1).data()
     }
+    // Multi-label classification is not supported
   }
 
   async assess (data: data.Data, useConfusionMatrix: boolean = false): Promise<Array<{ groundTruth: number, pred: number, features: Features }>> {
     const batchSize = this.task.trainingInformation?.batchSize
     if (batchSize === undefined) {
       throw new TypeError('Batch size is undefined')
     }
-
     const model = await this.getModel()
 
     let features: Features[] = []
@@ -52,7 +51,7 @@ export class Validator {
           const xs = e.xs as tf.Tensor
           const ys = await this.getLabel(e.ys as tf.Tensor)
           const pred = await this.getLabel(await model.predict(xs))
-
+          
           const currentFeatures = await xs.array()
           if (Array.isArray(currentFeatures)) {
             features = features.concat(currentFeatures)
@@ -70,7 +69,7 @@ export class Validator {
           throw new Error('Input data is missing a feature or the label')
         }
       }).toArray()).flat()
-
+    
     this.logger.success(`Obtained validation accuracy of ${this.accuracy}`)
     this.logger.success(`Visited ${this.visitedSamples} samples`)
 

docs/examples/README.md

@@ -43,3 +43,7 @@ You can run the custom task example with:
 cd docs/examples
 npm run custom_task # compiles TypeScript and runs custom_task.ts
 ```
+
+### Creating a CSV file to connect a dataset in DISCO
+
+DISCO allows connecting data through a CSV file that maps filenames to labels. The python notebook `dataset_csv_creation.ipynb` shows how to create such a CSV and the `scin_dataset.ipynb` shows how to download the [SCIN dataset](https://github.com/google-research-datasets/scin/tree/main?tab=readme-ov-file) and how to preprocess it in a format accepted by DISCO.