ONE-PEACE for Object Detection & Instance Segmentation

Pretrained Models

Name	batch size	iter	box AP	mask AP	download
onepeace_det	64	90k	60.4	52.9	model

Installation

git clone https://github.com/facebookresearch/detectron2.git
python -m pip install -e detectron2[all]

Evaluation

Object detection:

python lazyconfig_train_net.py --config-file ./configs/onepeace/cascade_mask_rcnn_vitdet_50ep.py --num-gpus 4 --eval-only train.init_checkpoint=/path/to/model_checkpoint

Expected results:

Task: bbox
AP,AP50,AP75,APs,APm,APl
60.3704,79.2223,65.5209,44.8737,64.9417,75.5091

Instance segmentation:

python lazyconfig_train_net.py --config-file ./configs/onepeace/cascade_mask_rcnn_vitdet_50ep.py --num-gpus 4 --eval-only train.init_checkpoint=/path/to/model_checkpoint model.roi_heads.maskness_thresh=0.5

Expected results:

Task: segm
AP,AP50,AP75,APs,APm,APl
52.8513,77.0620,58.0247,34.1789,56.3485,71.3040

Training

python lazyconfig_train_net.py --config-file ./configs/onepeace/cascade_mask_rcnn_vitdet_50ep.py --num-gpus 8 --num-machines 8 --machine-rank ${RANK} --dist-url "tcp://$MASTER_ADDR:60900" model.backbone.net.pretrained=${CHECKPOINT_PATH} train.output_dir=${OUTPUT_DIR}