nschang/DeepPlastic

Detecting and Quantifying Marine Debris using Deep Visual Models.

Deep Plastic

• Enhanced Object Detection for Epipelagic Plastic.
• This repository contains source code for the method developed in DeepPlastic: Identifying Marine Plastic In The Epipelagic Zone using Computer Vision and Deep Learning

Information:

• Paper: [Coming Soon]
• Preprint: https://arxiv.org/
• YouTube video of Results: https://youtu.be/8zBdFxaK4Os
• Data: Google Drive

Object Detection Model

• Two models: YOLOv4 and YOLOv5
• Small efficient and high precision models can be used for real-time object detection.
• Model architecture and implementation details: https://arxiv.org/
• Weights for YOLOv4 and YOLOv5 are provided in the model/
  • YOLOv4: best.weights; use best.weights
  • YOLOv5: best.pt; use best.pt

Google Colab Links

Note: Click on File and Save Copy in Drive. If you try to edit my file it'll ask you for permissions and send me an email. Please make your own copy.

• YOLOv5: https://colab.research.google.com/drive/1_qzbpBWkNfxQ0ny-DvsKicCeM0aFU4eW?usp=sharing

DeepTrash DataSet

• 1900 training images, 637 test images, 637 validation images (60, 20, 20 split)
• Field images taken from Lake Tahoe, San Francisco Bay and Bodega Bay in CA.
• Internet images (<20%) taken by scraping Google Images.
• Deep Sea images are from JAMSTEK JEDI dataset: http://www.godac.jamstec.go.jp/
• Complete DeepTrash dataset can be downloaded from: Google Drive

Results

Bibliography entry:

@misc{tata2021deepplastic,
  title={DeepPlastic: A Novel Approach to Detecting Epipelagic Bound Plastic Using Deep Visual Models}, 
  author={Gautam Tata and Sarah-Jeanne Royer and Olivier Poirion and Jay Lowe},
  year={2021},
  eprint={2105.01882},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}

Feeling Generous? Support our cause on Paypal