Focus On Wildlife CMP Scripts

Scripts for processing Focus On Wildlife -- Cleveland Metroparks classification data

Forked from similar scripts at Snapshot Serengeti Web link to GitHub repo[1]

These scripts will take the raw data produced by the citizen science project Focus On Wildlife -- Cleveland Metroparks and turn them into processed datasets that can then be used for analysis.

Notes from Ali Swanson on how to get started duplicating the aggregate report

• Brooke has some code at

  • Code for running basic aggregation on the raw data
  • imports two scripts from the main Data-digging directory

• The official aggregation approach and code for Snapshot Serengeti (e.g. www.nature.com/articles/sdata201526) is here

• The Oxford team is working to fix aggregation through the project builder - I'm hopeful we might be able to have something built this summer, but there are no guarantees.

Zooniverse report types

• workflow contents - has workflow id, contents version, and question choices
• workflows - has workflow id, workflow version, tasks, and some other stuff I don't understand fully
• subjects - has subject, project, workflow, subject_set IDs, image names, image links, classification count, retirement date/time, reason
  • this is the one to look at to see what has not been retired yet
  • could be used to see what is currently linked to workflow
• classifications - included user information and ip, workflow_id and version, created at date/time, session information, classification answers, subject data and ids

Basic generic reports

• These should be executed in a Canopy command prompt window
  • in Canopy command shell, cd \Users\pdl\Documents\GitHub\FocusOnWildlifeCMPScripts)
  • then execute the commands listed below

basic_project_stats.py

python basic_project_stats.py focus-on-wildlife-cleveland-metroparks-classifications_test.csv workflow_id=1432 workflow_version=478.99 --keep_allcols outfile_csv=focus-on-wildlife-cleveland-metroparks-classifications_wfid1432_v478.99.csv --remove_duplicates --keep_nonlive

Produces output like this:

Computing project stats using: infile: focus-on-wildlife-cleveland-metroparks-classifications_test.csv Reading classifications from focus-on-wildlife-cleveland-metroparks-classifications_test.csv Considering only workflow id 1432 Considering only major workflow version 478 Retaining all non-live classifications in analysis. Found 1227 duplicate classifications (0.55 percent of total). Duplicates removed from analysis (1125 unique user-subject-workflow groups).

Overall:

221497 classifications of 25396 subjects by 3158 classifiers, 1920 logged in and 1238 not logged in, from 3708 unique IP addresses. 189231 classifications were from logged-in users, 32266 from not-logged-in users.

That's 8.72 classifications per subject on average (median = 8.0). The most classified subject has 18 classifications; the least-classified subject has 1.

Median number of classifications per user: 18.00 Mean number of classifications per user: 70.14

Top 10 most prolific classifiers: user_name barclem 15347 martyreynard 6382 Giant_Sloth 5790 jfelix2 3921 jgeschke 2952 apmabr 2661 jrhansler 2206 auforester 1926 Zandra 1724 ramekens 1628 Name: created_at, dtype: int64

Gini coefficient for classifications by user: 0.77

Classifications were collected between 2017-01-05 and 2017-05-25. The highest classification id considered here is 56718037.

File with used subset of classification info written to focus-on-wildlife-cleveland-metroparks-classifications_wfid1432_v478.99.csv . File with ranked list of user classification counts written to focus-on-wildlife-cleveland-metroparks-classifications_test_nclass_byuser_ranked.csv . Saved info for all classifications that have duplicates to focus-on-wildlife-cleveland-metroparks-classifications_test_duplicated_only.csv .

• We can use this output file to run R scripts or postgres to summarize IDs and calculate Pielou scores, etc.

Modified aggregation report - this is not currently too useful

• Brooke has modified her aggregation report and put it here: Weblink

python aggregate_cleveland_survey.py focus-on-wildlife-cleveland-metroparks-classifications_wfid1432_v478.99.csv

Produces:

236894 annotations jailbroken from 221497 classifications, written to focus-on-wildlife-cleveland-metroparks-classifications_wfid1432_v478.99_annotations_1lineeach.csv as individual marks...

Aggregated classifications written to focus-on-wildlife-cleveland-metroparks-classifications_wfid1432_v478.99_aggregated.csv (kitchen sink version: focus-on-wildlife-cleveland-metroparks-classifications_wfid1432_v478.99_aggregated_kitchensink.csv )

• This approach mangles answers to our deer questions, among other problems

Plot histogram of user classifications

python plot_user_class_hist.py focus-on-wildlife-cleveland-metroparks-classifications_test_nclass_byuser_ranked.csv

Output was some errors and 'Histograms written to nclass_hist.png and .pdf'

Plot time series

python active_users_timeseries.py focus-on-wildlife-cleveland-metroparks-classifications_wfid1432_v478.99.csv workflow_id=1432 workflow_version=478 focus-on-wildlife-cleveland-metroparks-classifications_wfid1432_v478.99

User session stats

python sessions_inproj_byuser.py focus-on-wildlife-cleveland-metroparks-classifications_wfid1432_v478.99.csv

Outputs to session_stats_2017-01-05_to_2017-05-25.csv

Todo

• Make sense of aggregate report
• Get Peilou score etc. from Ali's scripts

[1] A description of all Snapshot Serengeti approach and datasets (raw dataset and processed datasets can be found here: doi:10.5061/dryad.5pt92), the field methods, and uses for the data can be found in: Swanson et al. 2015. Snapshot Serengeti - High-resolution annotated camera trap images of 40 mammal species in Serengeti National Park. Scientific Data.