/object_detection

Primary LanguageJupyter Notebook

object_detection

https://tensorflow-object-detection-api-tutorial.readthedocs.io/en/latest/training.html

https://towardsdatascience.com/creating-your-own-object-detector-ad69dda69c85

https://github.com/datitran/raccoon_dataset

Gather Images, split train/test = 80/20 or 90/10

train, file1-file80

test, file81-file100

python transform_image_resolution.py -d images/train/ -s 800 600

python transform_image_resolution.py -d images/test/ -s 800 600

ImgLabel https://github.com/tzutalin/labelImg

cd ~/labelImg

In data/predefined_classes.txt define the list of classes that will be used for your training.

python labelImg.py

python xml_to_csv.py

This creates two files in the images directory:

test_labels.csv

train_labels.csv

Before we can transform the newly created files to TFRecords we need to change a few lines in the generate_tfrecords.py file.

def class_text_to_int(row_label):
    if row_label == 'Raspberry_Pi_3':
        return 1
    elif row_label == 'Arduino_Nano':
        return 2
    elif row_label == 'ESP8266':
        return 3
    elif row_label == 'Heltec_ESP32_Lora':
        return 4
    else:
        return None

Now the TFRecords can be generated by typing:

python generate_tfrecord.py --csv_input=data/train_labels.csv --image_dir=images/train --output_path=train.record

python generate_tfrecord.py --csv_input=data/test_labels.csv --image_dir=images/test --output_path=test.record

Creating a label map

The label map maps an id to a name. We will put it in a folder called training, which is located in the object_detection directory. The labelmap for my detector can be seen below.

item {
    id: 1
    name: 'Steel_Pole'
}
item {
    id: 2
    name: 'Wood_Pole'
}

The id number of each item should match the id of specified in the generate_tfrecord.py file.

https://raw.githubusercontent.com/tensorflow/models/master/research/object_detection/samples/configs/faster_rcnn_inception_v2_pets.config

https://raw.githubusercontent.com/tensorflow/models/master/research/object_detection/model_main.py

pip3 install -U scikit-image pip3 install -U cython pip3 install "git+https://github.com/philferriere/cocoapi.git#egg=pycocotools&subdirectory=PythonAPI"

tf_upgrade_v2
--infile model_main.py
--outfile model_main_v2.py

python model_main_v2.py --logtostderr --model_dir=training/ --pipeline_config_path=training/faster_rcnn_inception_v2_pets.config