/lstm_tracker

Simple LSTM Network for Object Tracking

Primary LanguagePythonMIT LicenseMIT

LSTSM Multi Object Tracking Tracking

Requirenments

  • Anaconda3/Miniconda3

  • Tensorflow 2.9

  • Pytorch (latest)

pip install -r requirements.txt

Training The LSTM Tracker

Put all the MOTA format .txt files for training in <PATH>/lstm_tracker/data/MOTA/

  • MOTA format => <frame_id> <track_id> <x_centre> <y_centre> <w> <h> <_obj_class>

Open a Terminal and Navigate to the training folder

cd <PATH>/lstm_tracker/trainer/
chmod +x train.sh #(only needed the first time)
train.sh <PATH>/lstm_tracker/

Running an Inference with the Tracker

Download KITTI Object Tracking Video Sequences and Save all the sequences from 0000 to 0020 in <PATH>/lstm_tracker/tracker_testing/data/

cd <PATH>/lstm_tracker/tracker/
chmod +x run_tracker.sh #(only needed the first time)
run_tracker.sh <PATH>/lstm_tracker/ <PATH>/lstm_tracker/tracker_testing/data/<FOLDER> <INT>

  • <INT> is an integer corresponding to <FOLDER> sequence.

  • <FOLDER> should have frame wise txt file in a sub folder named labels

Real Time Tracking

  • Open a Terminal window
cd <PATH>/lstm_tracker/yolo/
python detect.py --project <PATH>/lstm_tracker/tracker_testing/ --source <PATH>/lstm_tracker/tracker_testing/data/<FOLDER> --save-txt
  • Here <FOLDER> is the directory which contains the videos in form of frames.
  • Open a new Terminal window
cd <PATH>/lstm_tracker/tracker/
chmod +x run_tracker.sh #(only needed the first time)
run_tracker.sh <PATH>/lstm_tracker/ <PATH>/lstm_tracker/tracker_testing/data/<FOLDER> -1
  • <FOLDER> is same as in the previous command

Results

  • The video with GT boxes is stored in <PATH>/lstm_tracker/tracker_testing/exp/FOLDER and GT boxes coordinates are stored in <PATH>/lstm_tracker/tracker_testing/exp/FOLDER/labels/ (in a frame wise .txt file format)
  • Results of MOT are stored in <PATH>/lstm_tracker/tracker_testing/results/FOLDER.txt

Evaluation

  • Evaluation for KITTI Test Dataset
  • Open a Terminal window
cd <PATH>/lstm_tracker
chmod +x track_eval.sh #(only needed for the first time)
track_eval.sh <PATH>/lstm_tracker/
  • The Evaluation Result is displayed in the Terminal window