/locating-objects-without-bboxes

PyTorch code for locating objects without bounding boxes - Loss function and trained models

Primary LanguagePythonOtherNOASSERTION

Locating Objects Without Bounding Boxes

PyTorch code for https://arxiv.org/pdf/1806.07564.pdf

Citing this work

@article{ribera2019,
  title={Locating Objects Without Bounding Boxes},
  author={Javier Ribera and David G\"{u}era and Yuhao Chen and Edward J. Delp},
  journal={Proceedings of the Computer Vision and Pattern Recognition (CVPR)},
  month={June},
  year={2019},
  note={{Long Beach, CA}}
}

Datasets

The datasets used in the paper can be downloaded from:

Installation

Use conda to recreate the environment provided with the code:

conda env create -f environment.yml

Activate the environment:

conda activate object-locator

Install the tool:

pip install .

Usage

Activate the environment:

conda activate object-locator

Run this to get help (usage instructions):

python -m object-locator.locate -h
python -m object-locator.train -h

Example:

python -m object-locator.locate \
       --dataset DIRECTORY \
       --out DIRECTORY \
       --model CHECKPOINTS \
       --evaluate \
       --no-gpu \
       --radius 5
python -m object-locator.train \
       --train-dir TRAINING_DIRECTORY \
       --batch-size 32 \
       --env-name sorghum \
       --lr 1e-3 \
       --val-dir TRAINING_DIRECTORY \
       --optim Adam \
       --save saved_model.ckpt

Pre-trained models

Models are trained separately for each of the four datasets, as described in the paper:

  1. Mall dataset
  2. Pupil dataset
  3. Plant dataset
  4. ShanghaiTechB dataset

The COPYRIGHT of the pre-trained models is the same as in this repository.

Uninstall

conda deactivate object-locator
conda env remove --name object-locator

Code Versioning

The code used in the paper corresponds to the tag used-for-cvpr2019-submission. If you want to reproduce the results, checkout that tag with git checkout used-for-cvpr2019-submission. The master branch is the latest version available, with convenient bug fixes and better documentation. If you want to develop or retrain your models, we recommend the master branch. Versions numbers follow semantic versioning and the changelog is in CHANGELOG.md.