Click here to Skip to main content
15,947,476 members
Please Sign up or sign in to vote.
0.00/5 (No votes)
I trained a YOLO v5 object detection model on my custom dataset, then converted it to TensorFlow using the export function and to TensorFlow lite using tf.convertor.
I want to use the TFLite model in an android app, however, the problem is the file size is relatively large (~ 27.5 MB).
Is there a way to make its size smaller without affecting the accuracy?

What I have tried:

I read about quantization and pruning but I don't know what would be best to not affect my model's performance.
Member 15627495 22-Mar-23 12:11pm    
Please, give details about the content of your dataset ( type , length , depth , nature of content)

do you think about 'remote computation' ?
or compression when the datas are not processing ?
Fatema Shawki 23-Mar-23 8:39am    
The dataset consists of 4 classes, each class consists of .jpg files and .txt files (contains the locations and class of the object in yolo format)
I'm sorry but I'm currently unclear on the implications of what you are suggesting. Could you please provide more information or clarification so that I can better understand?

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

CodeProject, 20 Bay Street, 11th Floor Toronto, Ontario, Canada M5J 2N8 +1 (416) 849-8900