https://github.com/douglasrizzo/fruit_detection_interview

https://github.com/douglasrizzo/fruit_detection_interview

Last synced: 25 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/douglasrizzo/fruit_detection_interview
• Owner: douglasrizzo
• Created: 2021-06-24T08:48:26.000Z (over 3 years ago)
• Default Branch: master
• Last Pushed: 2024-03-13T23:24:16.000Z (11 months ago)
• Last Synced: 2024-12-17T19:28:28.746Z (about 1 month ago)
• Language: Jupyter Notebook
• Size: 88.9 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# 2021 ML and Com. Vis. Engineer, Problem 2

- Training and evaluation report available [here](https://wandb.ai/tetamusha/fruit_detection_torchvision/reports/Detecting-oranges-with-TorchVision--Vmlldzo4MDAyNzM)

- The Jupyter Notebook + PDF contain a dataset study, which allowed me to discover the number of object classes, object annotations, bounding box aspect ratios and size in pixels.

- The code is basically aa usage example of the code in [this other repository](https://github.com/douglasrizzo/fruit_detection).

The best model trained for the problem used the following arguments:

    python train.py --batch_size 4 --num_workers 4 --backbone mobilenet --detector fasterrcnn --augmentations imgaug --epochs 350 --lr_initial 0.0005 --lr_final 0.00001 --lr_updates 50 --eval_size 0.1 --img_min_size 1920 --img_max_size 1920