Google AI Open Images - Object Detection Track

https://www.kaggle.com/c/google-ai-open-images-object-detection-track

File description

• Training model （Faster and SSD）: train_frcnn.py
• Training model （Yolo）: train_yolo.py
• Generate the submission csv: eval.py
• Evaluate with yolo: yolo_eval_csv.py
• Convert raw img to tfrecord: create_oid_tf_record.py
• Visual check the submission: vis_infer_result.py
• Visual check tfrecord file: check_tfrecord.py

Working folder description

• script: the bash files to run all the app
• config: config files needed for app
• All other main files should placed in the root folder
• web: a simple framework to run a server app that can recogization simgle image.

Data file description

• tfrecord files:
• box.csv: ground truth of box and classes info for each image
• oid_object_detection_challenge_500_label_map.pbtxt: mapping between machine code of class and human readable name
• faster_rcnn.config: config files for tfrecord, checkpoint

Download dataset

• https://github.com/cvdfoundation/open-images-dataset
• https://docs.aws.amazon.com/cli/latest/userguide/installing.html
• https://storage.googleapis.com/openimages/web/download.html
  • Only download the csv from this site
  • The csv files include the box, image class, image URL information

Install Code

• sudo /home/chamo/.pyenv/versions/anaconda3-5.1.0/bin/protoc ./object_detection/protos/*.proto --python_out=.

Requirement

• tensorflow 1.8
• Other python package which can be eaily installed by pip

Members:

• https://www.kaggle.com/rabienrose
• https://www.kaggle.com/bryanbocao
• https://github.com/rabienrose
• https://github.com/BryanBo-Cao/x-lab

Links

• http://cv-tricks.com/object-detection/faster-r-cnn-yolo-ssd/
• https://towardsdatascience.com/understanding-ssd-multibox-real-time-object-detection-in-deep-learning-495ef744fab
• https://www.figure-eight.com/dataset/open-images-annotated-with-bounding-boxes/
• https://itnext.io/implementing-yolo-v3-in-tensorflow-tf-slim-c3c55ff59dbe