/Kaels-toolbox-1

Invoker Kael's toolbox, a collection of frequently used scripts.

Primary LanguagePython

Kaels-toolbox

Invoker Kael's toolbox, a collection of frequently used scripts.

Content

List of available tools

  1. caffe-cubicle - Caffe tools
    • caffe-visualizer - Draw curves according to training log
    • caffe-fm-visualizer - Visualize internal featuremap
    • img-cls - Image-classification task
  2. caffe2-cubicle - Caffe2 and Detectron tools
    • caffe2_setup.sh - Auto-install Caffe2
    • detectron_setup.sh - Auto-install Detectron based on caffe2
  3. labelX-cubicle - LabelX tools
    • gen_labelx_jsonlist.py - Generate labelX-standard jsonlist from raw data
    • labelx_jsonlist_adapter.py - Convert labelX-standard jsonlist to useable format
    • gen_ava_jsonlist.py - Generate Ava-standard jsonlist from raw data
    • ava_jsonlist_adapter.py - Convert Ava-standard jsonlist to useable format
  4. mxnet-cubicle - MXNet tools
    • feat-extra - Image Feature Extraction
    • img-cls - Image-classification task
    • mxnet-cam - Draw class activation mapping
    • mxnet-visualizer - Draw curves according to training log
    • obj-det - Object-detection task
    • mxnet_setup.sh - Auto-install mxnet
  5. pytorch-cubicle - Pytorch tools
  6. qwheel-cubicle - Q-wheels
    • domain-pop - Search for bucket name of specified domain
    • log-proxy - Export service logs of AtServing
    • qhash-proxy - Get remote file hash(md5/sha1) on bucket
  7. warden-cubicle - Open Dataset tools
    • openimage - openimage v4 data processing
  8. checkout_md5.py - Check hash(md5) of files and exclude repetitive ones
  9. classification_evaluator.py - Evaluate image classification results
  10. download_from_urls.py - Multi-threading downloading scripts
  11. filter_log_acc.sh - Grep epoch accuracy of training logs
  12. ss_download.sh - Download file from source-station to accelerate
  13. verify_fexist.sh - Verify whether files in path list are available

Requirements

  • Most of the scripts require docopt interface pack:

    pip install docopt

  • If No module named AvaLib is warned:

    # either remove this line and corresponding lines in codes
import AvaLib

# or install AvaLib with cmd
easy_install lib/AvaLib-1.0-py2.7.egg

Usage

use xxx -h to get help information