=====================observations.sqlite3=====================
The database observations.sqlite3 has been generated from looking at the photos included separately, and also contains other metadata (e.g. experiment days and locations).

The tables are as follows:

==experimentdays==
Records where the experimentation was taking place etc.

Columns:
date = date YYYY-MM-DD
description = some additional notes of the experiment that day.
real bees = whether real bees were tagged and tracked in this dataset
location & location2 = string describing the location(s). 
  edge field - field at North end of The Edge student village in Sheffield.
  endcliffe hall avenue - a road in Sheffield.
  edge minimeadow - a small meadow planting area by a large pond at the Edge student village, in Sheffield.
  besbury park garden - a garden at a house on Besbury Park in Minchinhampton, Glos.
  minch field - A field in Minchinhampton (51°41'57.6"N 2°11'09.2"W).
  chedworth window - A garden at a house in Chedworth, Glos. (photos taken from window of house)
  chedworth garden - A garden at a house in Chedworth, Glos. (photos taken from garden).
track = score from 0 to 100 of whether that data will provide good examples of tracking.
casestudy = score from 0 to 100 of whether the data for that day will provide a good single bee-track case study.
useful observations = notes about what worked well on that day. 

==found==
Bees that have been redetected.
beeid = the bee id found (we don't always manage to id the unique bee, e.g. if it flies away before we get a good photo).
flower = which flower (if any) the bee is on.
location = not recorded
datetime = day/time found
obsid = which observation photo set this refers to (these photosets are from standard-camera photos taken of the found bees, that are grouped into particular 'events' or 'sets'.
confidence = how confident we are in the id.
notes = notes

==labels==
The bees that have been tagged.
Not all the bees might have survived this.
Not all are redetected.

beeid = the id of the bee
siteid = where the bee was tagged
firstdatetime = when it tagged
species = the species of the bee tagged
caste = w,m,q - worker, male or queen
obsid = which observation images show this bee
confidence = how confident are we on the species.

==tageffect==
To investigate the effect of tags we recorded the time from seeing a bee to it leaving the forage (and its caste & plant)
tagged = whether the bee was tagged
plant = which plant (a mix of latin and common names)
species = the species of the bee
caste = the caste of the bee
date = the date seen
starttime, endtime = the time from when we first saw the bee to when it left the forage.
notes = notes.

=====================labelled.zip=====================
Are images of the tagged bees, the file names have a number, this refers to the number in the table.

=====================20200802.zip, 20200803.zip, 20200805.zip, flightpathdata.zip=====================
This archives contain many images taken by the tracking system. 20200802,20200803,20200805 contain the images for experiments looking at foraging preferences, while flightpathdata.zip contains the two sets of data from two tracking systems both looking at the same bee nest at the same time. This data was used to reconstruct the 3d flight path of a bee.

Access:
The files inside are all python3 numpy (pickled) files.
The filenames are of the form: 
photo_object_YYYYMMDD_hh:mm:ss.zzzzzz_idex.np
zzzzzz=millisecond
idex=index of the image
e.g. photo_object_20200802_19:04:46.714933_0365.np

Python code to open it:
import numpy as np
a = np.load('photo_object_20200802_19:04:46.714933_0365.np',allow_pickle=True)

Example structure of the file: It contains a dictionary, such as this one:
{'index': 365,
 'img': array([[7, 7, 6, ..., 0, 0, 0],
        [6, 6, 7, ..., 0, 0, 0],
        [6, 6, 6, ..., 0, 0, 0],
        ...,
        [0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0]], dtype=uint8),
 'record': {'direction': 0,
  'flash': False,
  'triggertimestring': '20200802_19:04:46.714933',
  'endofset': True,
  'triggertime': 1596391486.7149158,
  'index': 365,
  'flashselection': [0]}}

'index' = the index of the photo
'img' = unsigned 8bit integer greyscale image
'record' = a dictionary containing metadata about the experiment
  'direction' = unused, but in future will record the direction the camera was facing, 'flash' = whether the flash fired (typically the image times are in pairs with a flash & a no flash photo). 'triggertimestring' = the time the trigger was requested. 'triggertime' = same but in numbers. 'endofset' = a series of photos taken closely in time are collected. This flag is set true at the end of that set (this allows the onboard tracking system to know to start computing on the set). 'index' = same as other index. 'flashselection' = which of the four flashes was used.