https://github.com/bkj/tagless

Interface for building image classifiers, via transfer learning, active search and uncertainty sampling
https://github.com/bkj/tagless

active-learning active-search human-computer-interaction labeling-tool python rest-api

Last synced: 8 months ago
JSON representation

Interface for building image classifiers, via transfer learning, active search and uncertainty sampling

• Host: GitHub
• URL: https://github.com/bkj/tagless
• Owner: bkj
• License: mit
• Created: 2017-08-01T21:59:33.000Z (almost 9 years ago)
• Default Branch: master
• Last Pushed: 2023-03-08T20:10:18.000Z (over 3 years ago)
• Last Synced: 2023-08-01T10:38:07.180Z (almost 3 years ago)
• Topics: active-learning, active-search, human-computer-interaction, labeling-tool, python, rest-api
• Language: Python
• Homepage:
• Size: 2.06 MB
• Stars: 2
• Watchers: 3
• Forks: 1
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
## tagless

Tagging interface w/ transfer learning, linearized active search and uncertainty sampling:

|                   |                              | 

| ----------------- | ---------------------------- |

| Transfer learning | https://github.com/bkj/tdesc |

| Linearized active search (LAS) |  https://github.com/bkj/simple_las | 

| Uncertainty sampling | https://github.com/bkj/libact | 

Under active development -- some things are broken or don't have sensible APIs exposed.

### Usage

```

    cd $TARGET_DIR

    mkdir -p ./{data,results}

    # expect set of images to be in `imgs` directory

    

    # Featurize images

    find ./imgs/ -type f | python -m tdesc --model vgg16 --crow > .crow

    

    # Prep + reformat images

    python $TAGLESS_ROOT/tagless/prep.py --inpath .crow ./data/crow

    

    # Run server

    python -m tagless --outpath ./results/my-labels --crow ./data/crow

    

    # Connect to localhost:5000 + start tagging

```

### Notes

Uncertainty sampling computes the score for each unlabeled image at each iteration.  ATM we're using a linear SVM, so the runtime of this step increases linearly w/ the size of the corpus.  On my machine, predicting on ~350K images takes ~2.5s, which is unacceptably slow.  Thus, for big corpora, we may want to fall back to some kind of approximate matrix-vector product. That'll take a little bit of thought thought.  For now I'll recommend running on a subset of the data.

__Idea__: Feature vectors are normalized relus -- so norm=1 and all positive entries.  Could maybe do uncertainty sampling via `faiss` by using vector orthogonal to SVM feature vector and take the largest/smallest entries.  Have to check my work on that one though.

### Dependencies

This has been tested on Ubuntu 16.04 w/ Python 2.7 (via Anaconda)