• Create a list of all classes by image count
• Download images for custom lists of classes (using parallelization)
• Delete corrupt images
• Train a model of choice on the downloaded image dataset
• Evaluate the performance of the model (includes per-class accuracies)

Python 3.6 or higher

Other package versions may work too.
Can be installed from requirements.txt

1. Download the Image IDs, Image labels, Boxes and Class Names from https://storage.googleapis.com/openimages/web/download.html
   (Train, Validation and Test of "Subset with Image-Level Labels" and Bounding Boxes of "Subset with Bounding Boxes")

2. Put them in a folder structure like this:
   

3. Create folders named out and processing

4. Run the script 1_create_class_id_to_image_ids.py
   Output:
   

5. Run the script 2_create_class_list_by_image_count.py
   Output:
   

6. Choose class names to train your classifier on from out/class_list_by_image_count and put them into a .txt file inside in/class_lists
   Example:
   

7. Adjust all options in config.py under # image download to your liking

8. Run the script 3_download_images.py
   Example Output:
   

9. Run the script 4_delete_corrupt_images.py

10. Adjust all options in config.py under # model training to your liking

11. Run the script 5_train_model.py
    Output:
    
    Now you have an Tensorflow Image classifier at out/saved_model

12. If you killed the previous script because it took too long, run 6_extract_model_from_checkpoint.py

13. Run the script 7_evaluate_model.py
    Output:
    

14. DONE

• The dataset is very noisy, you might have to manually delete images that do not fit the label
• Make sure you have enabled GPU support https://www.tensorflow.org/install/gpu
• Place your dataset on a SSD drive (500Mb/s should be enough) for faster training