Object Detection Tools

This repository is useful tools for TensorFlow Object Detection API.

Only Demo

　For only demo. Setup Python3.x, TensorFlow 1.x or TensorFlow 2.x, OpenCV.

Then execute following commnads, you can get object detection demo on Mac/Linux PC/Jetson Nano/Raspberry Pi.

TensorFlow 2.x

$ cd && git clone https://github.com/karaage0703/object_detection_tools
$ cd ~/object_detection_tools/models
$ ./get_efficientdet_d0_coco17_tpu-32.sh
$ cd ~/object_detection_tools
$ python3 scripts/object_detection_tf2.py -l='./models/coco-labels-paper.txt' -m='./models/efficientdet_d0_coco17_tpu-32/saved_model/'

TensorFlow 1.x

$ cd && git clone https://github.com/karaage0703/object_detection_tools
$ cd ~/object_detection_tools/models
$ ./get_ssdlite_mobilenet_v2_coco_model.sh
$ cd ~/object_detection_tools
$ python3 scripts/object_detection.py -l='models/coco-labels-paper.txt' -m='models/ssdlite_mobilenet_v2_coco_2018_05_09/frozen_inference_graph.pb'

Setup

Setup Python3.x and TensorFlow environment.

And get TensorFlow Models repository.

Execute following commands for download TensorFlow Object Detection API and change directory:

$ git clone https://github.com/tensorflow/models
$ cd models/research

Go to models/research directory

Usage

Download this repository

Execute following command:

$ git clone https://github.com/karaage0703/object_detection_tools

Model download

Change directory object_detection_tools/models and execute download script for downloading model file.

For example:

$ ./get_ssd_inception_v2_coco_model.sh

Test Prediction

Execute following commands at object_detection_tools after downloading ssd_inception_v2_coco_model data:

$ cd ~/object_detection_tools
$ python scripts/object_detection.py -l='models/coco-labels-paper.txt' -m='models/ssd_inception_v2_coco_2018_01_28/frozen_inference_graph.pb'

Train

Annotate data

Using VoTT is recommended.

Export tfrecord data.

Convert tf record file name

Put tfrecord data ./data/train and ./data/val directory.

Then, execute following command at object_detection_tools/data directory:

$ ./change_tfrecord_filename.sh

Train Models

SSD inception v2 example(fine tuning)

Change directory object_detection_tools/models and execute download script for downloading model file:

$ ./get_ssd_inception_v2

Execute following commands for training model:

$ export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim
$ python object_detection/model_main.py --pipeline_config_path="./object_detection_tools/config/ssd_inception_v2_coco.config" --model_dir="./saved_model_01" --num_train_steps=1000 --alsologtostderr

notice: model_dir must be empty before training

Convert Model

Convert from ckpt to graph file.

Execute following commands for converting from ckpt to graph file:

$ export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim
$ python object_detection/export_inference_graph.py --input_type image_tensor --pipeline_config_path object_detection_tools/config/ssd_inception_v2_coco.config --trained_checkpoint_prefix saved_model_01/model.ckpt-1000 --output_directory exported_graphs

Convert Label

Convert from pbtxt data to label data.

Execute follwing commands for converting from pbtxt data to label data:

$ export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim
$ python object_detection_tools/scripts/convert_pbtxt_label.py -l='object_detection_tools/data/tf_labl_map.pbtxt' > ./exported_graphs/labels.txt

Test trained model

Execute following command for testing trained model:

$ python object_detection_tools/scripts/object_detection.py -l='./exported_graphs/labels.txt' -m='./exported_graphs/frozen_inference_graph.pb'

License

This software is released under the Apache 2.0 License, see LICENSE.

References

• https://github.com/tensorflow/models
• https://github.com/tsutof/tf-trt-ssd