A tool to annotate images in a large dataset efficiently by incorporating Machine Learning techniques.
We developed an offline mobile web app (local setup available) that is designed primarily to annotate unlabeled images. To minimize the cost of annotation (i.e number of images to be manually annotated), we use Batch Active Learning for predicting the labels of the next set of images by an ML model which learns from the images manually annotated by a domain expert.
- Fast Human Annotation
- Incorporating Active Learning
- File I/O consistency
- Clone the repo.
- Download Miniconda from here
https://docs.conda.io/en/latest/miniconda.html - Once downloaded open the exectuable File
Miniconda3-latest-Windows-x86_64. And follow the usual installation process. - After the installation gets completed open command-prompt and type
conda --version. If you get a prompt saying:conda 4.9.2, you have correctly installed. - Create a conda env:
conda create -y -n at37 python=3.7``` - Create an environment:
-
conda activate at37
-
- Navigate to the local folder where you have the cloned repo.
- Ensure you are present in Seeds_Project/Ann_Tool_Seeds_Proj
- Install dependencies
pip install -r requirements.txt
-
python create_start_state.py --is_os_win 0 --initials hk --run 1 --global_reset 0 --img_dir_path ./static/Path2ImageFolder
-
python main.py --is_os_win 0 --initials hk --img_dir_path ./static/Path2ImageFolder
- Copy everything after 'Dash is running on' say (http://127.0.0.1:7236) and open a new browser tab (say Chrome/Mozilla etc) and paste in the URL field of the tab.