FlowVerify

This is the source code for the CoRL-19 paper:

Combining Deep Learning and Verification for Precise Object Instance Detection
Siddharth Ancha*, Junyu Nan*, David Held
Carnegie Mellon University

Project Website: https://jnan1.github.io/FlowVerify

Usage:

To setup this repo:

cd FlowVerify

export PYTHONPATH=/path/to/FlowVerify

To train FlowMatchNet:

cd FlowVerify/flowmatch/exps/coco

python main.py --config=fn_cc/whole/config

To run FlowMatchNet on detector outputs and generate score files:

cd FlowVerify/pipeline/eval

python run_flowverify.py --config=../configs/tdid_rgbd.yaml --model-path=/path/to/pretrained/model --out-dir=/path/to/output/dir --save-every=100

Check files in FlowVerify/pipeline/configs/ for example configuration files

To evaluate using FlowVerify:

cd FlowVerify/combine

python eval_params.py --config=config.yaml --eval_ds=gmu_test

Make sure to change paths in FlowVerify/combine/config.yaml to point to desired files