diff --git a/src/constants/machineLearning/config.ts b/src/constants/machineLearning/config.ts
index 454599fae..536ec425a 100644
--- a/src/constants/machineLearning/config.ts
+++ b/src/constants/machineLearning/config.ts
@@ -90,4 +90,4 @@ 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];
+export const SCENE_DETECTION_IMAGE_SIZE = 224;
diff --git a/src/services/machineLearning/imageSceneService.ts b/src/services/machineLearning/imageSceneService.ts
index b48448630..21ae8949a 100644
--- a/src/services/machineLearning/imageSceneService.ts
+++ b/src/services/machineLearning/imageSceneService.ts
@@ -7,6 +7,7 @@ import {
     Versioned,
 } from 'types/machineLearning';
 import { SCENE_DETECTION_IMAGE_SIZE } from 'constants/machineLearning/config';
+import { resizeToSquare } from 'utils/image';
 
 class ImageScene implements SceneDetectionService {
     method: Versioned<SceneDetectionMethod>;
@@ -54,24 +55,16 @@ class ImageScene implements SceneDetectionService {
 
     async detectScenes(image: ImageBitmap, minScore: number) {
         await tf.ready();
+        // scene detection model takes fixed-shaped (224x224) inputs
+        // https://tfhub.dev/sayannath/lite-model/image-scene/1
+        const resized = resizeToSquare(image, SCENE_DETECTION_IMAGE_SIZE);
 
         const model = await this.getImageSceneModel();
 
         const output = tf.tidy(() => {
-            const tensor = tf.browser.fromPixels(image);
-
-            // This model takes fixed-shaped (224x224) inputs
-            // https://tfhub.dev/sayannath/lite-model/image-scene/1
-            let resizedTensor = tf.image.resizeBilinear(
-                tensor,
-                SCENE_DETECTION_IMAGE_SIZE as [number, number]
-            );
-
-            resizedTensor = tf.expandDims(resizedTensor);
-            resizedTensor = tf.cast(resizedTensor, 'float32');
-
-            const output = model.predict(resizedTensor) as tf.Tensor;
-
+            const tfImage = tf.browser.fromPixels(resized.image);
+            const input = tf.expandDims(tf.cast(tfImage, 'float32'));
+            const output = model.predict(input) as tf.Tensor;
             return output;
         });