beenull/ente
1
0
Fork 0
Code Issues Pull requests Actions Packages Projects Releases Wiki Activity

refactor

Browse source
This commit is contained in:
Rushikesh Tote 2022-06-01 18:33:27 +05:30
parent b2acec63d8
commit 793cf9d5a1
2 changed files with 15 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/constants/machineLearning/config.ts
View file

@ -89,3 +89,5 @@ export const MOBILEFACENET_FACE_SIZE = 112;
export const TESSERACT_MIN_IMAGE_WIDTH = 44;
export const TESSERACT_MIN_IMAGE_HEIGHT = 20;
export const TESSERACT_MAX_IMAGE_DIMENSION = 720;
export const SCENE_DETECTION_IMAGE_SIZE = [224, 224];

20
src/services/machineLearning/imageSceneService.ts
View file

@ -6,6 +6,7 @@ import {
SceneDetectionService,
Versioned,
} from 'types/machineLearning';
import { SCENE_DETECTION_IMAGE_SIZE } from 'constants/machineLearning/config';
class ImageScene implements SceneDetectionService {
method: Versioned<SceneDetectionMethod>;
@ -35,9 +36,11 @@ class ImageScene implements SceneDetectionService {
this.model = model;
// warmup the model
const warmupResult = this.model.predict(tf.zeros([1, 224, 224, 3]));
await (warmupResult as tf.Tensor).data();
(warmupResult as tf.Tensor).dispose();
const warmupResult = this.model.predict(
tf.zeros([1, 224, 224, 3])
) as tf.Tensor;
await warmupResult.data();
warmupResult.dispose();
}
async detectScenes(image: ImageBitmap, minScore: number) {
@ -52,18 +55,21 @@ class ImageScene implements SceneDetectionService {
// This model takes fixed-shaped (224x224) inputs
// https://tfhub.dev/sayannath/lite-model/image-scene/1
let resizedTensor = tf.image.resizeBilinear(tensor, [224, 224]);
let resizedTensor = tf.image.resizeBilinear(
tensor,
SCENE_DETECTION_IMAGE_SIZE as [number, number]
);
resizedTensor = tf.expandDims(resizedTensor);
resizedTensor = tf.cast(resizedTensor, 'float32');
const output = this.model.predict(resizedTensor);
const output = this.model.predict(resizedTensor) as tf.Tensor;
return output;
});
const data = await (output as tf.Tensor).data();
(output as tf.Tensor).dispose();
const data = await output.data();
output.dispose();
const scenes = this.getScenes(
data as Float32Array,