Skip to content

Latest commit

 

History

History
212 lines (151 loc) · 5.53 KB

CHANGELOG.md

File metadata and controls

212 lines (151 loc) · 5.53 KB
Raw

Changelog

Date format: YY/MM/DD

2020/08/15

  • add atom online module from pytracking & DROL
  • update speed defination in vot test

2020/08/08

  • adopt pre-commit to format code while committing
  • reformat code with isort, autoflake and yapf

2020/07/30

  • support amp from pytorch>=1.6

2020/06/29

  • support acclerating tracking by tensorRT

2020/06/23

  • refine make_densebox_target.py
    • target making code: numpy -> pytorch
    • simplified for SOT task
      • reduce to single resolution layer
  • relax restrictions in requirements.txt
    • e.g. version of pytorch, opencv, numpy, etc.

2020/05/08

  • format evaluation code
  • Long-term tracking
    • add dataset and test_engine for UAV123 and VOTLT
    • add test results for UAV123
    • add long-term trakcker (need to be refined)

2020/04/28

  • Modify interface of siamfcpp_track.py: update
    • Accept provided target state (bbox)

2020/04/25

  • Add multi-processing tester
    • otb
    • lasot
    • trackingnet
    • got10k

2020/04/18

  • add SAT train code
  • update SAT test code
  • change TrackInfo to TextInfo
  • new param loader for model borrowed from dectron2
  • update coco dataset from detectron2

2020/04/18

  • New features in SOT demo (sot_video.py)
    • Add image file video reader (reading from image files)
    • Add image file video dumping (dump result as frame-level image files)

2020/04/13

  • Add OTB test

2020/04/12

  • Add iterable dataset in
    • dataset_type under config node: data
      • regular: former dataset
      • iterable: newly added dataset
    • data/adaptor_dataset.py: AdaptorIterableDataset / MultiStreamDataloader
    • data/builder.py: the branch with cfg.dataset_type
  • Fix seed issue
    • pass builder's seed to dataset as ext_seed

2020/04/06

  • Add model/pipeline/tester for SAT on DAVIS
  • Add transferred models for SAT

2020/03/25

Major changes

  • Add hyper-parameter search script
    • main/hpo.py
    • experiment
    • videoanalyst/utils/hpo.py

2020/03/23

Major changes

  • Add full training data
    • coco / det / vid / trackingnet / lasot / got10k
    • with caching mechanism (reducing storage I/O by caching indexes into memory)
    • TODO: perform verification with full data (under batchsize=256)

2020/03/25

Major changes

  • Add hyper-parameter search script
    • main/hpo.py
    • experiment
    • videoanalyst/utils/hpo.py

2020/03/23

Major changes

  • Add full training data
    • coco / det / vid / trackingnet / lasot / got10k
    • with caching mechanism (reducing storage I/O by caching indexes into memory)
    • TODO: perform verification with full data (under batchsize=256)

2020/03/18

Major changes

  • Add caching in LaSOT dataset
  • Add multi-mode in siamfcpp visualization debugging tools
    • demo/main/debug/visualize_siamfcpp_training_data.py
    • Seperate config file for debugging: experiments/siamfcpp/data/siamfcpp_data-trn.yaml

2020/03/17

Major changes

  • Update all training configuration
    • the cuda device number is defined by num_processes rather than devices in training phase
    • the settings between DP and DDP are unified now
  • Adopt loguru to replace logging for better logger
  • Update code for distributed training
  • Update code for model/pipeline/trainer

2020/03/14

Major changes

  • Update all training configuration
    • learning rate multiplied by 2 in order to be compatible with PyTorch==1.4.0 & CUDA==10.1
    • Update requirements.txt
  • Add tinyconv training configuration
    • experiments/siamfcpp/train/siamfcpp_tinyconv-trn.yaml
  • Add dataloader visualization tools for SiamFC++ (SOT)
    • demo/main/debug/visualize_siamfcpp_training_data.py

2020/03/07

Major changes

  • Add webcam demo
    • demo/main/video/sot_video.py

2020/03/02

Major changes

  • Add training with PyTorch Distributed Data Parallel (DDP)
    • main/dist_trian.py
    • .../trainer_impl/distributed_regular_trainer.py

2020/02/27

Major changes

  • Add TensorboardLogger in monitor
    • Recursively put engine_data as scalar into Tensorboard

2020/02/22

Major changes

  • Caching mechanism added to got10k dataset
    • Build cache at root_dir/subset.pkl (default), subset=[train, val, test]
    • Cache created/loaded in GOT10k.data_dict and will be queried every time __getitem__ is called
  • Remove num_iterations item in training .yaml files
    • replaced by value calculated with nr_image_per_epoch and minibatch
      • num_iterations = nr_image_per_epoch // minibatch
    • all training configs has been updated.

Minor changes

  • pipeline.builder's builder method has been changed
    • build_pipeline -> build

2020/02/15

  • Add contrib module's template
    • docs/TEMPLATES/contrib_module

2020/02/14

  • Add one-shot detection demo
  • Complete ShuffleNetV2x1.0 experiment

2020/02/13

  • Support Training and Test on GOT-10k for Single Object Tracking
    • SiamFC++-AlexNet
    • SiamFC++-GoogLeNet
    • SiamFC++-ShuffleNetV2
  • Training Components
    • Dataset helper, data sampler, transformer (data augmentation), etc.
    • Optimization: learing rate scheduler, dynamic freezing, etc.
    • Trainer & Monitor

2019/12/31

  • Support SiamFC++ test on VOT benchmark
    • SiamFC++-AlexNet
    • SiamFC++-GoogLeNet