Train-an-Object-Detection-Classifier

##windows 10 . anaconda python3.6 . using anaconda prompt

youtube tutorial

1 - Gathering images

in my case I use google images to find [chinook] photos , I gather 200 sample

2 - Convert extensoins

convert all images extensoins to xx.jpg , using Format Factory http://www.pcfreetime.com/

3 - Rename all images

go to images directory in my case chinook and type in cmd python renameFiles.py

4 - Label images

using labelImg https://github.com/tzutalin/labelImg unzip labelImg
run cmd and go to labelImg dir

conda install pyqt=5 
pyrcc5 -o resources.py resources.qrc
python labelImg.py

5 - Split images manually randomly

all my samples = 200 ,130 for train ,70 for test

6 - Installing TensorFlow-GPU

pip install --upgrade tensorflow-gpu

7 - Creat virtual environment

conda create -n tensorflow1 pip python=3.5 
activate tensorflow1 
pip install --ignore-installed --upgrade tensorflow-gpu

other necessary packages

(tensorflow1) C:\> conda install -c anaconda protobuf 
(tensorflow1) C:\> pip install pillow 
(tensorflow1) C:\> pip install lxml 
(tensorflow1) C:\> pip install Cython 
(tensorflow1) C:\> pip install jupyter 
(tensorflow1) C:\> pip install matplotlib 
(tensorflow1) C:\> pip install pandas 
(tensorflow1) C:\> pip install opencv-python 

8 - Download the full TensorFlow object detection repository

https://github.com/tensorflow/models.git

9 - Download faster_rcnn_inception_v2_coco

https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md

10 - Download my Repository

https://github.com/seraj94ai/Train-Object-Detection-Classifier.git unzip folder and copy paste in C:\tensorflow1\models\research\object_detection open cmd

cd C:\tensorflow1\models\research\object_detection
mkdir images
mkdir inference_graph
mkdir training

>>>>>>>> NOW just follow my instruction in note.txt

11 - Configure environment variable

Configure PYTHONPATH environment variable set PYTHONPATH=D:\AI\object-detection\tf_api\workspace\Train-Object-Detection-Classifier-master\models;D:\AI\object-detection\tf_api\workspace\Train-Object-Detection-Classifier-master\models\research;D:\AI\object-detection\tf_api\workspace\Train-Object-Detection-Classifier-master\models\research\slim

12 - Compile Protobufs

Protobuf (Protocol Buffers) libraries must be compiled , it used by TensorFlow to configure model and training parameters Open Anaconda Prompt and go to C:\tensorflow1\models\research

cd /d D:\AI\object-detection\tf_api\workspace\Train-Object-Detection-Classifier-master\models\research
protoc --python_out=. .\object_detection\protos\anchor_generator.proto .\object_detection\protos\argmax_matcher.proto .\object_detection\protos\bipartite_matcher.proto .\object_detection\protos\box_coder.proto .\object_detection\protos\box_predictor.proto .\object_detection\protos\eval.proto .\object_detection\protos\faster_rcnn.proto .\object_detection\protos\faster_rcnn_box_coder.proto .\object_detection\protos\grid_anchor_generator.proto .\object_detection\protos\hyperparams.proto .\object_detection\protos\image_resizer.proto .\object_detection\protos\input_reader.proto .\object_detection\protos\losses.proto .\object_detection\protos\matcher.proto .\object_detection\protos\mean_stddev_box_coder.proto .\object_detection\protos\model.proto .\object_detection\protos\optimizer.proto .\object_detection\protos\pipeline.proto .\object_detection\protos\post_processing.proto .\object_detection\protos\preprocessor.proto .\object_detection\protos\region_similarity_calculator.proto .\object_detection\protos\square_box_coder.proto .\object_detection\protos\ssd.proto .\object_detection\protos\ssd_anchor_generator.proto .\object_detection\protos\string_int_label_map.proto .\object_detection\protos\train.proto .\object_detection\protos\keypoint_box_coder.proto .\object_detection\protos\multiscale_anchor_generator.proto .\object_detection\protos\graph_rewriter.proto .\object_detection\protos\calibration.proto

just on time

cd /d D:\AI\object-detection\tf_api\workspace\Train-Object-Detection-Classifier-master\models\research
python setup.py build

python setup.py install

13 - Test TensorFlow setup

Test TensorFlow setup to verify it works

cd /d D:\AI\object-detection\tf_api\workspace\Train-Object-Detection-Classifier-master\models\research\object_detection

jupyter notebook object_detection_tutorial.ipynb

14 - Generate Training Data

TFRecords is an input data to the TensorFlow training model creat .csv files from .xml files

cd /d D:\AI\object-detection\tf_api\workspace\Train-Object-Detection-Classifier-master\models\research\object_detection
python xml_to_csv.py

This creates a train_labels.csv and test_labels.csv file in the \object_detection\images folder.

15 - Edit generate_tfrecord.py

edit generate_tfrecord.py and put your classes names

python generate_tfrecord.py --csv_input=images\train_labels.csv --image_dir=images\train --output_path=train.record
python generate_tfrecord.py --csv_input=images\test_labels.csv --image_dir=images\test --output_path=test.record

16 - Create a label map and edit the training configuration file.

go to \data copy pet_label_map.pbtxt to \training dir and rename it to labelmap.pbtxt

edit it to your class chinook

17 - Configure object detection tranning pipeline

cd C:\tensorflow1\models\research\object_detection\samples\configs copy faster_rcnn_inception_v2_pets.config past it in \training dir and edit it

a -

In the model section change num_classes to number of different classes

b -

fine_tune_checkpoint : C:/tensorflow1/models/research/object_detection/faster_rcnn_inception_v2_coco_2018_01_28/model.ckpt

c -

In the train_input_reader section change input_path and label_map_path as :
Input_path : C:/tensorflow1/models/research/object_detection/train.record
Label_map_path: C:/tensorflow1/models/research/object_detection/training/labelmap.pbtxt

d -

In the eval_config section change num_examples as :
Num_examples = number of files in \images\test directory.

e -

In the eval_input_reader section change input_path and label_map_path as :
Input_path : C:/tensorflow1/models/research/object_detection/test.record
Label_map_path: C:/tensorflow1/models/research/object_detection/training/labelmap.pbtxt

17.5 -test all files is ready

in training folder we have a folder that is result of extracting downloaded model e.g ssd_inception_v2_coco_2018_01_28 label_map.pbtxt sth like cofiguration file

18 - Run the Training (with ssd)

cd /d D:\AI\object-detection\tf_api\workspace\Train-Object-Detection-Classifier-master\models\research\object_detection
python train.py --logtostderr --train_dir=training/dir --pipeline_config_path=training/ssd_inception_v2_coco.config

19 - Tensorboard

in cmd type cd /d D:\AI\object-detection\tf_api\workspace\Train-Object-Detection-Classifier-master\models\research\object_detection tensorboard --logdir=training

20 - Export Inference Graph

training is complete ,the last step is to generate the frozen inference graph (.pb file) change “XXXX” in “model.ckpt-XXXX” should be replaced with the highest-numbered .ckpt file in the training folder:

clean inference_graph folder in object detection

python export_inference_graph.py --input_type image_tensor --pipeline_config_path training//ssd_inception_v2_coco.config --trained_checkpoint_prefix training/dir/model.ckpt-2000 --output_directory inference_graph

Appendix: Common Errors

1. ModuleNotFoundError: No module named 'deployment'

You can use “echo %PATH%” and “echo %PYTHONPATH%” to check the environment variables and make sure they are set up correctly. Also, make sure you have run these commands from the \models\research directory:

setup.py build
setup.py install

2. ImportError: cannot import name 'preprocessor_pb2'

ImportError: cannot import name 'string_int_label_map_pb2' (or similar errors with other pb2 files) This occurs when the protobuf files (in this case, preprocessor.proto) have not been compiled go to step 12

3.

in C:\tensorflow1\models\research\object_detection\utils\learning_schedules.py

replace range(num_boundaries) to [i for i in range(num_boundaries)]

(tensorflow1) C:\tensorflow1\models\research> python setup.py build
(tensorflow1) C:\tensorflow1\models\research> python setup.py install

4.

in C:\tensorflow1\models\research\object_detection\generate_tfrecord.py replace [else: None] by [else: return 0]