Installation

This has been tested on Ubuntu 16.04 python 2.7 environment.

Source code

git clone https://github.com/wonheeML/mtl-ssl-detection.git

Requirements

# from mtl-ssl-detection/
pip install --upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.7.0-cp27-none-linux_x86_64.whl
pip install -r requirements.txt

Check the protoc version >= 3.5.1 [protoc]

protoc --version

# if < 3.5.1,
PROTOC_ZIP=protoc-3.5.1-linux-x86_64.zip
curl -OL https://github.com/google/protobuf/releases/download/v3.5.1/$PROTOC_ZIP
sudo unzip -o $PROTOC_ZIP -d /usr/local bin/protoc
rm -f $PROTOC_ZIP
sudo chmod +rx /usr/local/bin/protoc

Protobuf Compilation

# from from mtl-ssl-detection/
protoc object_detection/protos/*.proto --python_out=.

Add Libraries to PYTHONPATH

# from from mtl-ssl-detection/
export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim

Note: This command needs to run from every new terminal you start. If you wish to avoid running this manually, you can add it as a new line to the end of your ~/.bashrc file.

Testing the Installation

You can test that you have correctly installed the Tensorflow Object Detection
API by running the following command:

# from from mtl-ssl-detection/
python object_detection/builders/model_builder_test.py

Dataset

• MS COCO
  

• PASCAL VOC
  

Model Zoo

tensorflow object detection API provide a collection of detection models pre-trained on the ImageNet. The following models were tested.

# from from mtl-ssl-detection/object_detection/checkpoints/detection_model_zoo/
wget http://download.tensorflow.org/models/resnet_v1_101_2016_08_28.tar.gz
wget http://download.tensorflow.org/models/inception_resnet_v2_2016_08_30.tar.gz
wget http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_1.0_224.tgz
tar -xf resnet_v1_101_2016_08_28.tar.gz
tar -xf inception_resnet_v2_2016_08_30.tar.gz 
mkdir mobilenet_v1_1.0_224
tar -xvf mobilenet_v1_1.0_224.tgz -C mobilenet_v1_1.0_224