AGC - Nuvi_RecycleNet3

Effective trash detection model for AI Grand Challenge 2020 track 3 round 3.
Nuvilab Solution. The classes to be detected are:

• paper
• paper_pack
• steel
• glass
• PET
• plastic
• plasticbag

Usage:

Installation

cd mmvc
MMCV_WITH_OPS=1 pip install -e .
pip install mmdet==2.6.0

Get model weights

wget https://luca-ai-models.s3.ap-northeast-2.amazonaws.com/recyclenet3/model_checkpoint.pth -P ./model/

Run inference

python main.py --img_path ./sample_img.jpg

Will return a dictionary with the following format:

{'Annotations': [{'Label': 'steel', 'Bbox': [X1, Y1, X2, Y2], 'Confidence': 0~1}]}

The bbox format follows: [xmin, ymin, xmax, ymax]