Skip to content

sahilparekh/coco2017-tfrecord

Repository files navigation

coco2017-tfrecord

Create selective class tfrecord from coco2017 dataset very simple selective class tfrecord creator from coco2017 dataset.

The program works jupyter notebook (.ipynb file)

Getting Started

  1. Clone this Repo .
  2. Download cocoa2017 dataset with annotations files from: http://cocodataset.org/#download

Under "Images" download

2017 Train images [118K/18GB]
2017 Val images [5K/1GB]

Under "Annotations" download

2017 Train/Val annotations [241MB]

Folder Structure

  • <#Dataset path#>
    • annotations (folder)
    • train2017
    • val2017
    • test2017 (not compulsory)

Prerequisites

  1. Python 3
  2. COCO API
  3. Tensorflow
  4. numpy
  5. PILLOW
  6. Jupyter Notebook
1. Python 3 Installation

This you would already know

2. COCO API Installation (Only if the import statement in jupyter notebook does not work)

You will need COCO API. Installation instruction can be found on this link COCO API Still for your quick reference will list installation instruction:

Clone git repo:

git clone https://github.com/cocodataset/cocoapi
cd cocoapi/PythonAPI
make

*if you use 'python3' to run python files please make chanes in cocoapi/PythonAPI/Makefile (replace 'python' with 'python3')

3. tensorflow Installation
pip3 install tensorflow

or

pip3 install tensorflow-gpu
4. numpy Installation
pip3 install numpy
5. PILLOW Installation
pip3 install pillow
6. Jupyter notebook Installation
pip3 install jupyter

Running the program

Post cloning the Repo, go to repo dir.

jupyter notebook

the notebook will open in a browser. double click on ipynb file and start executing cell by cell :)

License

This project is licensed under the MIT License - see the LICENSE file for details

Star the REPO, if you find it useful. Feel free for pull requests.

CHEERS!!!