ROAD-R Challenge

This repository contains baseline code for the first and second tasks of the ROAD-R Challenge. The code is built on top of 3D-RetinaNet for ROAD.

The first task requires developing models for scenarios where only little annotated data is available at training time. More precisely, only 3 out of 15 videos (from the training partition train_1 of the ROAD-R dataset) are used for training the models in this task. The videos' ids are: 2014-07-14-14-49-50_stereo_centre_01, 2015-02-03-19-43-11_stereo_centre_04, and 2015-02-24-12-32-19_stereo_centre_04.

The second task requires that the models' predictions are compliant with the 243 requirements provided in constraints/requirements.txt.

Table of Contents

• Dependencies and data preparation
• Training
• Testing
• Post-processing

Dependencies and data preparation

For the dataset preparation and packages required to train the models, please see the Requirements section from 3D-RetinaNet for ROAD.

To download the pretrained weights, please see the end of the Performance section from 3D-RetinaNet for ROAD.

Training

To train the model, provide the following positional arguments:

• DATA_ROOT: path to a directory in which road can be found, containing road_test_v1.0.json, road_trainval_v1.0.json, and directories rgb-images and videos.
• SAVE_ROOT: path to a directory in which the experiments (e.g. checkpoints, training logs) will be saved.
• MODEL_PATH: path to the directory containing the weights for the chosen backbone (e.g. resnet50RCGRU.pth).

Example train command (to be run from the root of this repository):

DATASET="${HOME}/datasets/"
EXPDIR="${HOME}/experiments/ROAD-R_Challenge_SSL/"
KINETICS="${HOME}/experiments/kinetics-pt/"

mode=train
max_epochs=150
milestones="130,145"
lr=0.0041
batch_size=4

tnorm=Godel
req_weight=10

unlabelled_proportion=0.10
agentness_threshold=0.25

python main.py ${DATASET} ${EXPDIR}/${EXP_ID}/ ${KINETICS} \
        --MODE=$mode --MAX_EPOCHS=${max_epochs} --MILESTONES=$milestones \
        --LR=$lr --BATCH_SIZE=${batch_size} \
        --LOGIC=$tnorm --req_loss_weight=${req_weight} \
        --unlabelled_proportion=${unlabelled_proportion} --agentness_th=${agentness_th} 

Testing

Below is an example command to test a model.

CUDA_VISIBLE_DEVICES=1 python main.py /home/user/datasets /home/user/experiments/  /home/user/kinetics-pt/ --MODE=gen_dets --ARCH=resnet50 --MODEL_TYPE=I3D --DATASET=road --TRAIN_SUBSETS=train_1 --VAL_SUBSETS=test --SEQ_LEN=8 --TEST_SEQ_LEN=8 --BATCH_SIZE=4 --LR=0.0041 --NUM_WORKERS=8 --req_loss_weight=10.0 --LOGIC=Product 

This command will generate a file containing the detected boxes at the following location: /home/user/road/road/log-lo_cache_logic_<LOGIC>_<req_loss_weight>/<experiment-name>/detections-30-08-50_test/log-lo_ROAD_R_predictions_I3D_logic-Product-10.0.txt.