• Search by Module
• Search by Words
• Search Projects

• Java
• Python
• JavaScript
• TypeScript
• C++
• Scala
• Blog

Project: Pic-Numero (GitHub Link)

• Pic-Numero-master
  • Wheat_ROIs
  • Datasets
    • grain_glcm_d1_a4.data
    • grain_glcm_d1_a4_2.data
    • old_data.data
    • grain_glcm_d1_a4_2_new.data
  • dist
    • gui
      • _random.so
      • cStringIO.so
      • _collections.so
      • _multibytecodec.so
      • math.so
      • binascii.so
      • strop.so
      • zlib.so
      • _hashlib.so
      • itertools.so
      • resource.so
      • sip.so
      • qt_menu.nib
        • keyedobjects.nib
        • classes.nib
        • info.nib
      • select.so
      • operator.so
      • _heapq.so
      • array.so
      • datetime.so
      • _locale.so
      • fcntl.so
      • _struct.so
      • _functools.so
      • qt4_plugins
        • imageformats
          • libqgif.dylib
          • libqtga.dylib
          • libqsvg.dylib
          • libqico.dylib
        • codecs
          • libqkrcodecs.dylib
          • libqjpcodecs.dylib
          • libqtwcodecs.dylib
          • libqcncodecs.dylib
        • accessible
        • graphicssystems
          • libqglgraphicssystem.dylib
          • libqtracegraphicssystem.dylib
        • iconengines
          • libqsvgicon.dylib
      • cPickle.so
      • readline.so
      • _codecs_iso2022.so
      • bz2.so
      • time.so
  • Assets
  • LICENSE
  • test
  • Models
    • MLP_glcmdistance4.data
    • MLP.data
    • MLP_glcmdistance1.data
    • SVM_d1_a4_olddata.data
    • MLP_d1_a4.data
  • grain_images
  • PicNumero
    • PicNumero.pyc
    • build_classifier.py
    • PicNumero.py
    • CNN.pyc
    • RAG_threshold.py
    • spectral_roi.pyc
    • svm_visualise.py
    • blob.py
    • template_matching.py
    • Helper.py
    • gui_checkbox_handlers.py
    • spectral_roi.py
    • SVM.py
    • gui.py
    • glcm.pyc
    • Helper.pyc
    • Display.py
    • Display.pyc
    • glcm.py
    • MLP.pyc
    • MLP.py
    • tqdm
      • _tqdm_notebook.py
      • _version.pyc
      • _main.pyc
      • _tqdm_gui.py
      • _utils.pyc
      • _tqdm.pyc
      • __init__.pyc
      • _tqdm_pandas.pyc
      • _utils.py
      • __main__.py
      • _tqdm.py
      • _tqdm_gui.pyc
      • __init__.py
      • tests
        • tests_version.py
        • tests_tqdm.py
        • tests_pandas.py
        • tests_main.py
        • tests_perf.py
      • _version.py
      • _main.py
      • _tqdm_pandas.py
    • test.py
    • CNN.py
    • hook-skimage.io.py
    • SVM.pyc
    • count.py
  • Wheat_Images
  • README.md
  • train
    • negative
    • positive
  • test.py

Pic-Numero

This is a project exploring the counting of objects in images. While the project makes use of the example of counting wheat grains in images of wheat plants, it can be applied to objects of any kind. A very high-level description of the way the system works is that it trains a classifier to recognise images of the object being counted then attempts to recognise instances of this object in any given image and returns the number of recognised matches. As such, the count returned is really an estimate.

The system can make use of different classifiers depending on your choice (currently SVM, MLP neural net and CNN). It might be useful to know that the results of the classifications of the last count can be found in src/Results. "_0" means it was classified as false (not an object being counted) while "_1" means it was classified as true (object being counted).

Usage

First, to train the system to identify objects you want, all you have to do is populate the "train" folder (which contains "positive" and "negative" subfolders) with your training images appropriately.

Then in your code be sure to add from PicNumero import PicNumero.

You can then use PicNumero.run_with_svm() PicNumero.run_with_mlp() or PicNumero.run_with_cnn() with the filename of the image whose objects are to be counted as the argument.

from PicNumero import PicNumero

imagePath = "Path/To/Your/Image.png"
count = PicNumero.run_with_cnn(imagePath)

PicNumero also includes a helper standalone program for ROI extraction however it is only available for Mac OS X and certain flavours of Linux at this time. Can be found in dist/gui/. Run from terminal with the command "./dist/gui/gui". Source is contained in "gui.py" and "gui_checkbox_handlers.py".

Dependencies

Makes use of Python 2.7 and the scikit machine learning and image processing libraries (which can be found here and here).

Note: _TensorFlow is required to use the CNN functionality (CNN.py). If you do not have TensorFlow already installed, visit here for detailed instructions on how to install it._

Compatibility

Tested and working on OS X and certain flavours of Linux. Should work on Windows too but could potentially need some tweaking; I've never tried to run it on windows before.

Credits

Made use of Google's excellent TensorFlow library