Improved consensus UX

changelog.d/20260114_132218_klakhov_consensus_ux.md

@@ -0,0 +1,20 @@
+### Changed
+
+- Consensus merge function now preserves all shapes with their scores, regardless of quorum threshold
+  (<https://github.com/cvat-ai/cvat/pull/10172>)
+
+### Added
+
+- Score visualization in UI with a virtual "Votes" attribute calculated as `score × replica_jobs`
+  (<https://github.com/cvat-ai/cvat/pull/10172>)
+- A user now may navigate between different shapes with shortcuts (Tab/Shift+Tab by default) in Standard, Review modes
+  (<https://github.com/cvat-ai/cvat/pull/10172>)
+- Review mode now supports editing objects. Users can unlock and edit individual annotations as needed
+  (<https://github.com/cvat-ai/cvat/pull/10172>)
+- Double-clicking an object in the sidebar now centers it on the canvas and expands its details
+  (<https://github.com/cvat-ai/cvat/pull/10172>)
+
+### Removed
+- Consensus quorum setting has been removed. All merged annotations are now kept with their consensus scores,
+  allowing users to filter results based on score thresholds instead
+  (<https://github.com/cvat-ai/cvat/pull/10172>)

cvat-canvas/src/scss/canvas.scss

@@ -62,6 +62,11 @@ polyline.cvat_shape_drawing_opacity {
     font-style: oblique 40deg;
 }
 
+.cvat_canvas_text_score {
+    fill: #FFB347;
+    font-style: oblique 10deg;
+}
+
 .cvat_canvas_crosshair {
     stroke: red;
 }

cvat-canvas/src/typescript/canvasView.ts

@@ -3129,7 +3129,7 @@ export class CanvasViewImpl implements CanvasView, Listener {
         }
     }
 
-    private addText(state: any, options: { textContent?: string } = {}): SVG.Text {
+    private addText(state: any, options: { textContent?: string, isSkeletonElement?: boolean } = {}): SVG.Text {
         const { undefinedAttrValue } = this.configuration;
         const content = options.textContent || this.configuration.textContent;
         const withID = content.includes('id');
@@ -3138,10 +3138,14 @@ export class CanvasViewImpl implements CanvasView, Listener {
         const withSource = content.includes('source');
         const withDescriptions = content.includes('descriptions');
         const withDimensions = content.includes('dimensions');
+
         const textFontSize = this.configuration.textFontSize || 12;
         const {
-            label, clientID, attributes, source, descriptions,
+            label, clientID, attributes, source, descriptions, score, votes,
         } = state;
+        const isConsensus = source === 'consensus';
+        const withScore = isConsensus && !options.isSkeletonElement;
+        const withVotes = isConsensus && !options.isSkeletonElement;
 
         const attrNames = Object.fromEntries(state.label.attributes.map((attr) => [attr.id, attr.name]));
         if (state.shapeType === 'skeleton') {
@@ -3151,7 +3155,9 @@ export class CanvasViewImpl implements CanvasView, Listener {
                         textContent: [
                             ...(withLabel ? ['label'] : []),
                             ...(withAttr ? ['attributes'] : []),
+                            // Note: explicitly exclude 'score' and 'votes' for skeleton elements
                         ].join(',') || ' ',
+                        isSkeletonElement: true,
                     });
                 }
             });
@@ -3193,6 +3199,19 @@ export class CanvasViewImpl implements CanvasView, Listener {
                             .addClass('cvat_canvas_text_description');
                     });
                 }
+                if (withScore || withVotes) {
+                    const parts = [];
+                    if (withScore) {
+                        parts.push(`Score: ${score.toFixed(2)}`);
+                    }
+                    if (withVotes) {
+                        parts.push(`Votes: ${votes}`);
+                    }
+                    block
+                        .tspan(parts.join(', '))
+                        .attr({ dy: '1.25em', x: 0 })
+                        .addClass('cvat_canvas_text_score');
+                }
                 if (withAttr) {
                     Object.keys(attributes).forEach((attrID: string, idx: number) => {
                         const values = `${attributes[attrID] === undefinedAttrValue ?

cvat-core/src/annotations-collection.ts

@@ -68,6 +68,7 @@ export default class Collection {
         dimension: DimensionType;
         framesInfo: BasicInjection['framesInfo'];
         jobType: JobType;
+        consensusReplicas?: number;
     }) {
         this.stopFrame = data.stopFrame;
 
@@ -102,6 +103,7 @@ export default class Collection {
             nextClientID: () => ++config.globalObjectsCounter,
             getMasksOnFrame: (frame: number) => (this.shapes[frame] as MaskShape[])
                 .filter((object) => object instanceof MaskShape),
+            consensusReplicas: data.consensusReplicas,
         };
     }
 

cvat-core/src/annotations-filter.ts

@@ -80,6 +80,8 @@ interface ConvertedObjectData {
     type: ObjectType;
     shape: ShapeType;
     occluded: boolean;
+    score: number | null;
+    votes: number | null;
 }
 
 export default class AnnotationsFilter {
@@ -118,6 +120,8 @@ export default class AnnotationsFilter {
                 type: state.objectType,
                 shape: state.shapeType,
                 occluded: state.occluded,
+                score: state.score ?? null,
+                votes: state.votes ?? null,
             };
         });
 
@@ -166,6 +170,8 @@ export default class AnnotationsFilter {
                     shape: shape.type,
                     occluded: shape.occluded,
                     objectID: shape.clientID ?? null,
+                    score: shape.score ?? null,
+                    votes: null,
                 };
             }),
             tags: collection.tags.map((tag) => {
@@ -183,6 +189,8 @@ export default class AnnotationsFilter {
                     shape: null,
                     occluded: false,
                     objectID: tag.clientID ?? null,
+                    score: null,
+                    votes: null,
                 };
             }),
             tracks: collection.tracks.map((track) => {
@@ -200,6 +208,8 @@ export default class AnnotationsFilter {
                     shape: track.shapes[0]?.type ?? null,
                     occluded: null,
                     objectID: track.clientID ?? null,
+                    score: null,
+                    votes: null,
                 };
             }),
         };

cvat-core/src/annotations-objects.ts

@@ -55,6 +55,10 @@ function computeNewSource(currentSource: Source): Source {
         return Source.SEMI_AUTO;
     }
 
+    if (currentSource === Source.CONSENSUS) {
+        return Source.CONSENSUS;
+    }
+
     return Source.MANUAL;
 }
 
@@ -78,6 +82,7 @@ export interface BasicInjection {
     jobType: JobType;
     nextClientID: () => number;
     getMasksOnFrame: (frame: number) => MaskShape[];
+    consensusReplicas?: number;
 }
 
 type AnnotationInjection = BasicInjection & {
@@ -104,6 +109,8 @@ class Annotation {
     protected readOnlyFields: string[];
     protected color: string;
     protected source: Source;
+    public score: number;
+    public votes: number;
     public updated: number;
     public attributes: Record<number, string>;
     protected groupObject: {
@@ -127,6 +134,9 @@ class Annotation {
         this.readOnlyFields = injection.readOnlyFields || [];
         this.color = color;
         this.source = injection.jobType === JobType.GROUND_TRUTH ? Source.GT : data.source;
+        this.score = data.score;
+        this.votes = injection.consensusReplicas !== undefined ?
+            Math.round(this.score * injection.consensusReplicas) : 0;
         this.updated = Date.now();
         this.attributes = data.attributes.reduce((attributeAccumulator, attr) => {
             attributeAccumulator[attr.spec_id] = attr.value;
@@ -567,6 +577,7 @@ export class Shape extends Drawn {
             label_id: this.label.id,
             group: this.group,
             source: this.source,
+            score: this.score,
         };
 
         if (this.serverID !== null) {
@@ -606,6 +617,8 @@ export class Shape extends Drawn {
             pinned: this.pinned,
             frame,
             source: this.source,
+            score: this.score,
+            votes: this.votes,
             __internal: this.withContext(frame),
         };
 
@@ -956,6 +969,8 @@ export class Track extends Drawn {
             },
             frame,
             source: this.source,
+            score: this.score,
+            votes: this.votes,
             __internal: this.withContext(frame),
         };
     }
@@ -1489,6 +1504,8 @@ export class Tag extends Annotation {
             updated: this.updated,
             frame,
             source: this.source,
+            score: this.score,
+            votes: this.votes,
             __internal: this.withContext(frame),
         };
     }
@@ -2016,6 +2033,7 @@ export class SkeletonShape extends Shape {
             label_id: this.label.id,
             group: this.group,
             source: this.source,
+            score: this.score,
         };
 
         if (this.serverID !== null) {
@@ -2060,6 +2078,8 @@ export class SkeletonShape extends Shape {
             hidden: elements.every((el) => el.hidden),
             frame,
             source: this.source,
+            score: this.score,
+            votes: this.votes,
             __internal: this.withContext(frame),
         };
     }
@@ -3102,6 +3122,8 @@ export class SkeletonTrack extends Track {
             occluded: elements.every((el) => el.occluded),
             lock: elements.every((el) => el.lock),
             hidden: elements.every((el) => el.hidden),
+            score: this.score,
+            votes: this.votes,
             __internal: this.withContext(frame),
         };
     }

cvat-core/src/annotations.ts

@@ -97,6 +97,7 @@ async function getAnnotationsFromServer(session: Job | Task): Promise<void> {
             stopFrame: session instanceof Job ? session.stopFrame : session.size - 1,
             labels: session.labels,
             dimension: session.dimension,
+            consensusReplicas: session instanceof Job ? session.consensusReplicas : undefined,
             framesInfo: {
                 isFrameDeleted: session instanceof Job ?
                     (frame: number) => !!getJobFramesMetaSync(session.id).deletedFrames[frame] :

cvat-core/src/consensus-settings.ts

@@ -11,14 +11,12 @@ export default class ConsensusSettings {
     #id: number;
     #task: number;
     #iouThreshold: number;
-    #quorum: number;
     #descriptions: Record<string, string>;
 
     constructor(initialData: SerializedConsensusSettingsData) {
         this.#id = initialData.id;
         this.#task = initialData.task;
         this.#iouThreshold = initialData.iou_threshold;
-        this.#quorum = initialData.quorum;
         this.#descriptions = initialData.descriptions;
     }
 
@@ -38,14 +36,6 @@ export default class ConsensusSettings {
         this.#iouThreshold = newVal;
     }
 
-    get quorum(): number {
-        return this.#quorum;
-    }
-
-    set quorum(newVal: number) {
-        this.#quorum = newVal;
-    }
-
     get descriptions(): Record<string, string> {
         const descriptions: Record<string, string> = Object.keys(this.#descriptions).reduce((acc, key) => {
             const camelCaseKey = _.camelCase(key);
@@ -59,7 +49,6 @@ export default class ConsensusSettings {
     public toJSON(): SerializedConsensusSettingsData {
         const result: SerializedConsensusSettingsData = {
             iou_threshold: this.#iouThreshold,
-            quorum: this.#quorum,
         };
 
         return result;

cvat-core/src/enums.ts

@@ -89,6 +89,7 @@ export enum Source {
     AUTO = 'auto',
     FILE = 'file',
     GT = 'Ground truth',
+    CONSENSUS = 'consensus',
 }
 
 export enum EventScope {

cvat-core/src/object-state.ts

@@ -45,6 +45,8 @@ export interface SerializedData {
     color?: string;
     updated?: number;
     source?: Source;
+    score?: number;
+    votes?: number;
     zOrder?: number;
     points?: number[];
     occluded?: boolean;
@@ -88,6 +90,8 @@ export default class ObjectState {
         next: number | null;
         last: number | null;
     } | null;
+    public readonly score: number | undefined;
+    public readonly votes: number | undefined;
     public label: Label;
     public color: string;
     public hidden: boolean;
@@ -180,6 +184,9 @@ export default class ObjectState {
             objectType: serialized.objectType,
             shapeType: serialized.shapeType || null,
             updateFlags,
+            // score and votes are only available for shapes (not tracks or tags)
+            score: serialized.objectType === ObjectType.SHAPE ? (serialized.score ?? 1.0) : undefined,
+            votes: serialized.objectType === ObjectType.SHAPE ? (serialized.votes ?? 0) : undefined,
         };
 
         Object.defineProperties(
@@ -204,6 +211,12 @@ export default class ObjectState {
                 isGroundTruth: {
                     get: () => data.source === Source.GT,
                 },
+                score: {
+                    get: () => data.score,
+                },
+                votes: {
+                    get: () => data.votes,
+                },
                 clientID: {
                     get: () => data.clientID,
                 },

cvat-core/src/server-response-types.ts

@@ -377,7 +377,6 @@ export interface SerializedQualityReportData {
 export interface SerializedConsensusSettingsData {
     id?: number;
     task?: number;
-    quorum?: number;
     iou_threshold?: number;
     descriptions?: Record<string, string>;
 }
@@ -423,6 +422,7 @@ export interface SerializedShape {
     group: number;
     frame: number;
     source: Source;
+    score?: number;
     attributes: { spec_id: number; value: string }[];
     elements: Omit<SerializedShape, 'elements'>[];
     occluded: boolean;