Skip to content

sophiehsuu/tensorflow-object-detection-faster-rcnn

 
 

Repository files navigation

Object Detection with MobileNetSSDv2 and Faster R-CNN and the TensorFlow Object Detection API

Thannks to https://github.com/Tony607/object_detection_demo for the intial demo!

What is contained in this repo

  • A tutorial to train and use MobileNetSSDv2 with the TensorFlow Object Detection API
  • A tutorial to train and use Faster R-CNN with the TensorFlow Object Detection API

What you will learn (MobileNetSSDv2)

  • How to load your custom image detection from Roboflow (here we use a public blood cell dataset with tfrecord)
  • Download base MobileNetSSDv2 model
  • Set up training environment
  • Configure training pipeline and train the model
  • Export the trained model's .pb inference graph
  • Use the saved model for inference

What you will learn (Faster R-CNN)

  • How to load your custom image data from Roboflow (here we use a public blood cell dataset with tfrecord)
  • Download base pretrained Faster R-CNN model
  • Set up training environment
  • Configure training pipeline and train model
  • Export the trained model's .pb inference graph
  • Use the saved model for inference

Resources available

  • This blog post for MobileNetSSDv2 walks through the tutorial
  • This blog post for Faster R-CNN walks through the tutorial
  • For the MobileNetSSDv2 model tutorial Open In Colab
  • For the Faster R-CNN model tutorial Open In Colab
  • For reading purposes, for MobileNetSSDv2, the notebook is saved here as Tutorial_Mobilenet.ipynb
  • For reading purposes, for Faster R-CNN, the notebook is also saved here as Tutorial_Faster_RCNN.ipynb

About Roboflow for Data Management

Roboflow makes managing, preprocessing, augmenting, and versioning datasets for computer vision seamless. Developers reduce 50% of their code when using Roboflow's workflow, automate annotation quality assurance, save training time, and increase model reproducibility.

alt text

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 99.8%
  • Python 0.2%