Skip to main content

Utility functions for the fast ai mooc

Project description


Utility scripts for the online course. There are two main parts: one to download and organise arbitrary image classes, and one to highlight what parts of an image is activating the decision for a classification.

  1. Utility for Lesson 1 experimentation with external image classes. The script:
  • Downloads images from google images for specific classes
  • Sanity checks that images can be opened and have three channels
  • Organises the images into separate folders (train/valid/test + classes) as expected by the library
  1. Utility for creating Class Activation Maps for both classifications.


  • chromedriver is required. On ubuntu/debian: sudo apt-get chromium-chromedriver


pip install duckgoose


Fetching, sanity checking and organising images

from duckgoose import fetchImagesAndPrepForClassification

# dictionary structure `class_name => search term`
image_classes = { 'ducks' : 'ducks -rubber' , 'geese' : 'geese' }
download_path = '/home/myuser/data/downloaded_from_google'
output_path = '/home/myuser/data/ducksgeese/'
number_of_images = 100

fetchImagesAndPrepForClassification(image_classes, download_path, output_path, number_of_images)

Create Class Activation Maps (CAM)

Note: This was implemented for fastai v2 part 1. Here is a full example of creating a class activation maps for ducks and geese using fast ai.

from fastai.imports import *
from fastai.transforms import *
from fastai.conv_learner import *
from fastai.model import *
from fastai.dataset import *
from fastai.sgdr import *
from fastai.plots import *

from import calculateAndChartHeatZoneFor

PATH = "data/ducksgeese/"
arch = resnet34
bs = 64

m = arch(True)
m = nn.Sequential(*children(m)[:-2], 
                  nn.Conv2d(512, 2, 3, padding=1), 

tfms = tfms_from_model(arch, sz, aug_tfms=transforms_side_on, max_zoom=1.1)
data = ImageClassifierData.from_paths(PATH, tfms=tfms, bs=bs)
learn = ConvLearner.from_model_data(m, data)


_, val_tfms = tfms_from_model(learn.model, sz), 2)

calculateAndChartHeatZoneFor('./data/ducksgeese/test/ducks/ducks_427.jpg', val_tfms, learn)

Duck and goose heatmap


The MIT License (MIT)

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

duckgoose-0.1.8.tar.gz (5.2 kB view hashes)

Uploaded source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page