Scope

This repo is aiming to provide production ready 2D object detection code basics.

It's based on official tensorflow API jupyter notebook but I will gradually add more popular models such as yolo series.

More

If you are interested in 3D object detection, visit this repo.

If you are interested in Segmentation, visit this repo.

Introduction

This is a repo for implementing object detection with pre-trained models (as shown below) on tensorflow.

Model name	Speed	COCO mAP	Outputs
ssd_mobilenet_v1_coco	fast	21	Boxes
ssd_inception_v2_coco	fast	24	Boxes
rfcn_resnet101_coco	medium	30	Boxes
faster_rcnn_resnet101_coco	medium	32	Boxes
faster_rcnn_inception_resnet_v2_atrous_coco	slow	37	Boxes

Dependencies:

• Tensorflow >= 1.2.0
• OpenCV

Run Demo

# Clone this repo
git clone https://github.com/KleinYuan/tf-object-detection.git

# Setting up
cd tf-object-detection
bash setup.sh

# Run demo
python app.py

Image Classifications

I also put an image classifications inference app (VGG16) here.

# Assuming you already run setup.sh, which will download vgg16.np

python app_ic.py

Networks

Model name	Architecture
AlextNet
Vgg 16
SSD
ResNet
MobileNet
Faster R-CNN