azadehkhojandi / findingfishwithtensorflowobjectdetectionapi

Finding Fish with Tensorflow Object detection API

Step 1

On a GPU machine/VM go through steps in instalation.ipynb notebook. It should clone tensorflow models and does all the required setttings

Step 2

Copy fish_detection content inside models/research/object_detection

Step 3

Run the notebooks in the following order

1. object_detection_tutorial.ipynb
2. create_pascal_tf_record.ipynb
3. train.ipynb
4. fish_detection_tutorial.ipynb

• note if you get validation error for traing your model with fasterrcnn update def to_absolute_coordinates method and set check_range=False in models/research/object_detection/core/box_list_ops.py

`with tf.name_scope(scope, 'ToAbsoluteCoordinates'): height = tf.cast(height, tf.float32) width = tf.cast(width, tf.float32)

# Ensure range of input boxes is correct.
check_range=False`

you might need to rerun the following block to make sure your changes took place

import sys,os,os.path #manual step - hard coded issue ! cd ../ && {sys.executable} setup.py build && {sys.executable} setup.py install && chmod 777 slim && cd slim && ! sudo {sys.executable} -m pip install -e . && ! echo 'done' && ! ls

Resource:

1. Provision the Data Science Virtual Machine for Linux (Ubuntu)