Skip to content
View RE-OWOD's full-sized avatar

Block or report RE-OWOD

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
RE-OWOD/README.md

Revisiting Open World Object Detection [arXiv]

Installation

See INSTALL.md.

Dataset

Our new data division is based on COCO2017. We divide the training set into four tasks, in which each task has 20 categories. For each task, we obtained images containing the categories of each task from the training set, and removed the annotation information of other categories in these pictures during training. In each task, 1000 images are sampled as the validation set. And we de duplicate the training set and the validation set. For the testing set, we adopt the validation set of COCO2017, which contains relatively complete annotation information.

The data files are at ./datasets/Main/.

Train

Train with PAD:

The PAD module is implemented in ./detectron2/modeling/roi_heads/roi_heads.py,where the file generated by selective search needs to be loaded. The code that generates these files is in the ./datasets/save_selective_search.py.

Inference

Inference with CEC

python tools/train_net.py --num-gpus 3\
                          --eval-only --config-file ./configs/OWOD/t1/t1_test.yaml \
                          SOLVER.IMS_PER_BATCH 6 SOLVER.BASE_LR 0.005 \
                          OUTPUT_DIR "./output/t1/test_dir/" \
                          MODEL.WEIGHTS "/Path/To/Your/Model" \
                          OWOD.GENERATE_CALI False \
                          OWOD.CALI_PATH "/home/RE-OWOD-main/analysis/t1_REOWOD_train_scores_cali_07" \
                          OWOD.UNK_THRESH 0.6

Inference without CEC

python tools/train_net.py --num-gpus 3\
                          --eval-only --config-file ./configs/OWOD/t1/t1_test.yaml \
                          SOLVER.IMS_PER_BATCH 6 SOLVER.BASE_LR 0.005 \
                          OUTPUT_DIR "./output/t1/test_dir/" \
                          MODEL.WEIGHTS "/Path/To/Your/Model" \
                          OWOD.GENERATE_CALI True

If you want to generate CEC pickle file with train datasets, you need to repalce "./detectron2/evaluation/evaluator.py" with "./detectron2/evaluation/evaluator_cali.py". Remember to add current stage to CEC pickle files' name in "./detectron2/evaluation/evaluator_cali.py".

python tools/train_net.py --num-gpus 3\
                          --eval-only --config-file ./configs/OWOD/t1/t1_test.yaml \
                          SOLVER.IMS_PER_BATCH 6 SOLVER.BASE_LR 0.005 \
                          OUTPUT_DIR "./output/t1/test_dir/" \
                          MODEL.WEIGHTS "/Path/To/Your/Model" \
                          OWOD.GENERATE_CALI True

Popular repositories Loading

  1. RE-OWOD RE-OWOD Public

    revisiting open world object detection

    Python 77 7