Getting Started

This document provides tutorials to train and evaluate CenterNet. Before getting started, make sure you have finished installation and dataset setup.

COCO

To evaluate COCO object detection with HarDNet-85 run

python test.py ctdet --exp_id coco_h85 --arch hardnet_85 --load_model centernet_hardnet85_coco.pth

This will give an AP of 44.0 if setup correctly. The input images are resized to 512 x 512. You can add --flip_test and --flip_test --test_scales 0.5,0.75,1,1.25,1.5 to the above commend, for flip test and multi_scale test, respectively.

Training

We have packed all the training scripts in the experiments folder.

python main.py ctdet --exp_id coco_h85 --arch hardnet_85 --batch_size 48 --master_batch 24 --lr 1e-2 --gpus 0,1 --num_workers 16 --num_epochs 300 --lr_step 230,280