Open animesh-007 opened 3 years ago
The instance segmentation is built upon mmdetection framework v0.6.0. To run the code the mmdet package has to be built. The best is to create a virtual enironment with pytorch (at the time of development it was v1.1.0). The package is installed by these instructions:
pip install mmcv==0.2.8
cd mmdetection
./compile.sh
python setup.py develop
The config files for training are adapted for COCO dataset which has to be placed in data/ folder as described in mmdetection
./tools/dist_train.sh ${CONFIG_FILE} ${NUM_GPUS} [--validate]
Can you guide me, how to use harmonic network for instance segmentation using this code?