rchopinw / csci1430_final_project Goto Github PK

Brown University CSCI1430 Computer Vision Final Project: Real-time Face Mask Detection System

Contributors: Bangxi Xiao, Qiubai Yu, Xinyu Song, Tianqi Liu

Contact: [email protected] / [email protected] / [email protected] / [email protected]

FIRST THING FIRST: Install the required packages from requirements.txt

Usage (model preparation):

1. Cloning the git repository to local.
2. To download from GCP bucket, gsutil is needed. To install the gsutil tool, please check: https://cloud.google.com/storage/docs/gsutil_install (we highly recommend the pip method)
3. Afer installing the required tools, run "gsutil ls gs://csci1430-final" in command line to check the file structures in the bucket. To fetch the trained models, run "gsutil -m cp gs://csci1430-final/train_output/model {PATH_TO_THIS_REPO}/train_log".
4. (Alternative) You can also download model from our Google Drive: https://drive.google.com/drive/folders/1---YVlPH8-LfapZ6TiLRCcGOrpuJquam?usp=sharing

Usage (training, validation, and testing tfrecords data generation):

1. The original dataset for training and validation can be found via: https://www.kaggle.com/datasets/tapakah68/medical-masks-part1
2. The original testing data can be found via: https://www.kaggle.com/datasets/tapakah68/medical-masks-part7
3. After downloading and unzipping the files, place the images from 1 into data/train_images and images from 2 into data/test_images (according to ProcessARGS.py).
4. Run script convert_data.py and the tfrecords files will be automatically written into data/train_validation_data and data/test_data.
5. (Alternative) If you don't want to start from raw images, feel free to download the prepared tfrecords from our public GCP bucket: gs://csci1430-final/data

Usage (model training):

1. After data preparation, you can train the models by running script "train_model.py".
2. To train the model, you can use command line "python train_model.py --help" to learn more about the available options.
3. (Alternative) You can also fetch the models directly from our GCP bucket: gs://csci1430-final/train_output/model
4. For the training logs, they will be written into train_output/log, separated by different models. (You can also obtain the training logs from the bucket: gs://csci1430-final/train_output/log)

Usage (detection system):

1. Make sure that your PC has a camera.
2. Running script "run_detection.py". You might also want to learn more about the options we provide via "python run_detection.py --help".