CODEWITHSUNDEEP
pythontensorflowkerasobject-detectiondata-augmentation
Tiz
Tiz

Reputation: 51

Tensorflow: data augmentation with bounding box (for object detection) advice

I need advice for an effective data augmentation for an object detection problem (YOLO neural network) in tensorflow/keras. My first option was to use tf keras layers, but not all of them support bounding boxes. For example: keras_cv.layers.JitteredResize() supports bounding boxes but tf.keras.layers.RandomZoom() does not. What would be your advice? Additionally, I would be interested in the possibility of using ImageDataGenerator, is there a way to include my bounding boxes into the generator? Thank you a lot.

Upvotes: 1

Views: 141

Answers (1)

LetsTryMySolution
LetsTryMySolution

Reputation: 56

You can use librairies such as Albumentations if you want to do data augmentation with bounding boxes.

Upvotes: 0

Related Questions