Box deficit detection models

report

You can see the model report here

get trained model and result

Download zip file from

https://drive.google.com/open?id=1sB-zyM0wzE-yyrVSwSUmqNEFScUnjYvQ

and unzip it into the project root. This step are required to infer with the trained models.

DataSet

I'm sorry that the dataset is not allowed to publish.

How to train?

prerequieste

• anadonda3

install dependencies

In the project root,

conda env create -f=environment.yml

execute train script

In the project root,

export PYTHONPATH=.
python scripts/train.py $MODEL_NAME

chose $MODEL_NAME from

• "resnet50_pretrained"
• "resnet50_pretrained_cosine"
• "resnet50_pretrained_cosine_pruned"

"resnet50_pretrained_cosine_pruned" model have to be trained after training "resnet50_pretrained_cosine".

These train results are saved under output dir"

How to Infer with trained model?

Use Inferrer Class in inference.py.

Please see inference_sample.py to see how to use it.