This document provides tutorials to train and evaluate RAM. Before getting started, make sure you have finished installation and dataset setup.
To test our pretrained model on the validation set of PD, download the model, copy it to $RAM_ROOT/models/
, and run
cd $RAM_ROOT/src
python test.py tracking --exp_id ram_pd --dataset pd_tracking --dataset_version val --track_thresh 0.4 --load_model ../models/ram_pd.pth --is_recurrent --input_len 16 --random_walk --rw_head_depth 2 --pool_kernel 3 --max_age 16 --local_rw_r 0.2 --stream_test --new_thresh 0.5 --sup_reg
This will give a Track mAP of 71.96
if set up correctly. You can append --debug 4
to the above command to visualize the predictions.
To test the tracking performance on the validation set of KITTI with our pretrained model, download the model, copy it to $RAM_ROOT/models/
, and run
python test.py tracking --exp_id ram_kittihalf --dataset kitti_tracking --dataset_version val_half --track_thresh 0.4 --load_model ../models/ram_kittihalf.pth --is_recurrent --input_len 16 --debug 4 --random_walk --rw_head_depth 2 --pool_kernel 3 --max_age 16 --local_rw_r 0.2 --stream_test --new_thresh 0.5 --sup_reg --max_out_age 4
To test the tracking performance on the test set of LA-CATER, download the model, copy it to $RAM_ROOT/models/
, and run
python test.py tracking --exp_id ram_lacater_stage2 --dataset la_cater --dataset_version train --track_thresh 0.4 --load_model ../models/ram_lacater_stage2.pth --is_recurrent --debug 4 --input_len 70 --num_gru_layers 1 --debug 4 --random_walk --rw_head_depth 2 --pool_kernel 1 --max_age 300 --rw_score_thresh 0.005 --local_rw_r 0.1 --new_thresh 0.5 --stream_test --sup_reg --trainval
To test the tracking performance on the test set of LA-CATER-Moving, download the model, copy it to $RAM_ROOT/models/
, and run
python test.py tracking --exp_id ram_lacater_moving_stage2 --dataset la_cater_moving --dataset_version train --track_thresh 0.4 --load_model ../models/ram_lacater_moving_stage2.pth --is_recurrent --debug 4 --input_len 70 --num_gru_layers 1 --debug 4 --random_walk --rw_head_depth 2 --pool_kernel 1 --max_age 300 --rw_score_thresh 0.005 --local_rw_r 0.1 --new_thresh 0.5 --stream_test --sup_reg --trainval
We have packed all the training scripts in the experiments folder.
Each model is trained on 8 Tesla V100 GPUs with 32GB of memory.
If the training is terminated before finishing, you can use the same command with --resume
to resume training. It will found the latest model with the same exp_id
.
All experiments rely on existing pretrained models, we provide the links to the corresponding models directly in the training scripts.