franchesoni / iadet

IAdet : The ultimate annotation tool for object detection

Object detection is all about bounding boxes. The IAdet tool enables users to teach the computer how to draw bounding boxes around a particular class of objects present in some images. As you annotate, a model is trained on the background and is used to provide predictions.

Installation

1. Clone the repo
2. With Python 3.10 run:

python -m venv env_iadet
source env_iadet/bin/activate
bash install.sh  # installing mmcv might take a long time

Usage

bash launch.sh DATA_DIR

Cite

If you find this project useful, cite our work:

@inproceedings{marchesoni2022iadet,
  title={IAdet: Human in the loop object detection},
  author={Marchesoni-Acland, Facciolo},
  booktitle={NeurIPS 2022 Workshop on Human in the Loop Learning},
  year={2022}
}

or check out the paper's code here: https://github.com/franchesoni/iadet_paper

Future

• Include semi-supervised learning https://mmdetection.readthedocs.io/en/v3.0.0rc0/user_guides/semi_det.html#configure-meanteacherhook
• Make GUI show the images in the center of the region (resize accroding to max width max depth)

iadet's Issues

Demo filenames do not correspond with demo dataset?

Just tried to load the default "figs" images and annotations using Windows. Keeps saying that filenames do not correspond with dataset. I see that in your annotations you have figs/000001 as a file name, and also the file path. Does this make sense? Have you encountered this issue in Windows?

Also, the source part did not work for Windows, I had to use: env_IAdet/Scripts/activate

Thanks!