aboutsummaryrefslogtreecommitdiff
path: root/src/client/views/collections/collectionFreeForm/ImageLabelBox.tsx
diff options
context:
space:
mode:
authorbobzel <zzzman@gmail.com>2025-04-21 13:48:58 -0400
committerbobzel <zzzman@gmail.com>2025-04-21 13:48:58 -0400
commit17e24e780b54f2f7015c0ca955c3aa5091bba19c (patch)
treeb13002c92d58cb52a02b46e4e1d578f1d57125f2 /src/client/views/collections/collectionFreeForm/ImageLabelBox.tsx
parent22a40443193320487c27ce02bd3f134d13cb7d65 (diff)
parent1f294ef4a171eec72a069a9503629eaf7975d983 (diff)
merged with master and cleaned up outpainting a bit.
Diffstat (limited to 'src/client/views/collections/collectionFreeForm/ImageLabelBox.tsx')
-rw-r--r--src/client/views/collections/collectionFreeForm/ImageLabelBox.tsx33
1 files changed, 15 insertions, 18 deletions
diff --git a/src/client/views/collections/collectionFreeForm/ImageLabelBox.tsx b/src/client/views/collections/collectionFreeForm/ImageLabelBox.tsx
index a3d9641da..ff9fb14e7 100644
--- a/src/client/views/collections/collectionFreeForm/ImageLabelBox.tsx
+++ b/src/client/views/collections/collectionFreeForm/ImageLabelBox.tsx
@@ -9,9 +9,8 @@ import React from 'react';
import { imageUrlToBase64 } from '../../../../ClientUtils';
import { Utils, numberRange } from '../../../../Utils';
import { Doc, NumListCast, Opt } from '../../../../fields/Doc';
-import { DocData } from '../../../../fields/DocSymbols';
import { List } from '../../../../fields/List';
-import { ImageCast } from '../../../../fields/Types';
+import { ImageCastToNameType, ImageCastWithSuffix } from '../../../../fields/Types';
import { gptGetEmbedding, gptImageLabel } from '../../../apis/gpt/GPT';
import { DocumentType } from '../../../documents/DocumentTypes';
import { Docs } from '../../../documents/Documents';
@@ -165,9 +164,9 @@ export class ImageLabelBox extends ViewBoxBaseComponent<FieldViewProps>() {
// Converts the images into a Base64 format, afterwhich the information is sent to GPT to label them.
const imageInfos = this._selectedImages.map(async doc => {
- if (!doc[DocData].tags_chat) {
- const [name, type] = ImageCast(doc[Doc.LayoutFieldKey(doc)]).url.href.split('.');
- return imageUrlToBase64(`${name}_o.${type}`).then(hrefBase64 =>
+ if (!doc.$tags_chat) {
+ const url = ImageCastWithSuffix(doc[Doc.LayoutDataKey(doc)], '_o') ?? '';
+ return imageUrlToBase64(url).then(hrefBase64 =>
!hrefBase64 ? undefined :
gptImageLabel(hrefBase64,'Give three labels to describe this image.').then(labels =>
({ doc, labels }))) ; // prettier-ignore
@@ -176,7 +175,7 @@ export class ImageLabelBox extends ViewBoxBaseComponent<FieldViewProps>() {
(await Promise.all(imageInfos)).forEach(imageInfo => {
if (imageInfo) {
- imageInfo.doc[DocData].tags_chat = (imageInfo.doc[DocData].tags_chat as List<string>) ?? new List<string>();
+ imageInfo.doc.$tags_chat = (imageInfo.doc.$tags_chat as List<string>) ?? new List<string>();
const labels = imageInfo.labels.split('\n');
labels.forEach(label => {
@@ -186,7 +185,7 @@ export class ImageLabelBox extends ViewBoxBaseComponent<FieldViewProps>() {
.replace(/^\d+\.\s*|-|f\*/, '')
.replace(/^#/, '')
.trim();
- (imageInfo.doc[DocData].tags_chat as List<string>).push(hashLabel);
+ (imageInfo.doc.$tags_chat as List<string>).push(hashLabel);
});
}
});
@@ -200,13 +199,11 @@ export class ImageLabelBox extends ViewBoxBaseComponent<FieldViewProps>() {
groupImagesInBox = action(async () => {
this.startLoading();
- for (const doc of this._selectedImages) {
- for (let index = 0; index < (doc[DocData].tags_chat as List<string>).length; index++) {
- const label = (doc[DocData].tags_chat as List<string>)[index];
- const embedding = await gptGetEmbedding(label);
- doc[DocData][`tags_embedding_${index + 1}`] = new List<number>(embedding);
- }
- }
+ await Promise.all(
+ this._selectedImages
+ .map(doc => ({ doc, labels: doc.$tags_chat as List<string> }))
+ .map(({ doc, labels }) => labels.map((label, index) => gptGetEmbedding(label).then(embedding => (doc[`$tags_embedding_${index + 1}`] = new List<number>(embedding)))))
+ );
const labelToEmbedding = new Map<string, number[]>();
// Create embeddings for the labels.
@@ -215,13 +212,13 @@ export class ImageLabelBox extends ViewBoxBaseComponent<FieldViewProps>() {
// For each image, loop through the labels, and calculate similarity. Associate it with the
// most similar one.
this._selectedImages.forEach(doc => {
- const embedLists = numberRange((doc[DocData].tags_chat as List<string>).length).map(n => Array.from(NumListCast(doc[DocData][`tags_embedding_${n + 1}`])));
+ const embedLists = numberRange((doc.$tags_chat as List<string>).length).map(n => Array.from(NumListCast(doc[`$tags_embedding_${n + 1}`])));
const bestEmbedScore = (embedding: Opt<number[]>) => Math.max(...embedLists.map(l => (embedding && similarity(Array.from(embedding), l)!) || 0));
const {label: mostSimilarLabelCollect} =
this._labelGroups.map(label => ({ label, similarityScore: bestEmbedScore(labelToEmbedding.get(label)) }))
.reduce((prev, cur) => cur.similarityScore < 0.3 || cur.similarityScore <= prev.similarityScore ? prev: cur,
{ label: '', similarityScore: 0, }); // prettier-ignore
- doc[DocData].data_label = mostSimilarLabelCollect; // The label most similar to the image's contents.
+ doc.$data_label = mostSimilarLabelCollect; // The label most similar to the image's contents.
});
this.endLoading();
@@ -313,7 +310,7 @@ export class ImageLabelBox extends ViewBoxBaseComponent<FieldViewProps>() {
{this._displayImageInformation ? (
<div className="image-information-list">
{this._selectedImages.map(doc => {
- const [name, type] = ImageCast(doc[Doc.LayoutFieldKey(doc)]).url.href.split('.');
+ const [name, type] = ImageCastToNameType(doc[Doc.LayoutDataKey(doc)]);
return (
<div className="image-information" style={{ borderColor: SettingsManager.userColor }} key={Utils.GenerateGuid()}>
<img
@@ -322,7 +319,7 @@ export class ImageLabelBox extends ViewBoxBaseComponent<FieldViewProps>() {
await DocumentView.showDocument(doc, { willZoomCentered: true });
}}></img>
<div className="image-information-labels" onClick={() => this._props.addDocTab(doc, OpenWhere.addRightKeyvalue)}>
- {(doc[DocData].tags_chat as List<string>).map(label => {
+ {(doc.$tags_chat as List<string>).map(label => {
return (
<div key={Utils.GenerateGuid()} className="image-label" style={{ backgroundColor: SettingsManager.userVariantColor, borderColor: SettingsManager.userColor }}>
{label}