Kacper Trebacz
Jan Gruszczynski

Documentation

Yolact

First please run: pip install -r requirements.txt

Eval

If you want to run our models please type:

Fo resnet 50 model:

python src/models/yolact/eval.py --trained_model=models/yolact/best_model_adam.pth --config=yolact_resnet50_config --test

For darknet model:

python src/models/yolact/eval.py --trained_model=models/yolact/best_model53.pth --config=yolact_darknet53_config --test

if you want to see the results add --display argument

Train

If you want to train model on your own please type on of the following:

• dvc repro train_yolact
• dvc repro train_yolact

API

If you want to ru API build docker image:

docker build . -t yoloapi:1

Than run it:

docker run -p 8000:8000 yoloapi:1

And test with:

python src/api/test_script.py

Mask-RCNN

Resnet backbone:

Go to notebooks/mask-rcnn-4-resnet.ipynb
Training: Specify for how many epoch you want to train the network in the main function, then run all cells
If it comes to detection, follow the scheme in the last cell.

Mobilenetv2 backbone:

Training: Go to notebooks/mask-rcnn-4-resnet.ipynb
Specify for how many epoch you want to train the network in the main function, then run all cells
If it comes to detection, follow the scheme in the last cell.