Custom Object Detection Model Is Very Slow With Tensorflow.js

[Liju_Alivenow]

I have created my own object detection model using Google Colab and flowed through the tutorial given below.

How to Train TensorFlow Lite Object Detection Models Using Google Colab | SSD MobileNet

Around 100 images were labeled with Labellmg for single-object detection.

The chosen model is ssd-mobilenet-v2-fpnlite-320, and the generated model is Saved Model. We converted it into both the TFLite model and the Tensorflow.js graph model.

I followed the code from the tutorial given below.

Build Custom Object Detection Web Application Using TensorFlow.js

I was wondering how much faster COCO-SSD and Tensorflow.js are. It detects around 80 objects very quickly.

But my custom model (the TensorFlow.js graph model), which is trained for single-object detection, is freezing on browsers.

How can I improve my object detection speed on browsers, especially on mobile browsers?

[tagoma]

Hi @Liju_Alivenow. Please make sure to provide with as much as information as possible on your project so that people can help you. Thank you.