With TRACT you can assign attributes to images from a trail camera or similar site monitoring image source. TRACT reads EXIF data from the images for time tagging and allows groups of related images (same subjects, subject coming and going, etc.) to be defined to analyze activity durations.
The primary function on TRACT is to allow recording of relationships between images collected over time at a site. TRACT extracts EXIF data from images to get the time and date of image capture, and provides an interface that allows the user to relate multiple images together in a group. TRACT also allows entry of descriptors for images - composition of groups present, activity, etc., see the spreadsheet columns section for more details.
TRACT is intended to be a functional minimal viable product (MVP), solving the specific problem of relating images into groups and allowing associated data entry. Part of its minimalistic design is the use of a single Excel spreadsheet (.xlsx) as a data-store. The use of the Tkinter interface, a simple GUI library bundled with Python, also emphasizes a light weight low dependency approach. Similarly, although unrelated to the code is this repository, Auto PY to EXE is used to package TRACT into a zero install executable. This results in a large executable that is somewhat slow to launch, but avoids spending time on setting up a build environment for a more customized software stack.
- Distributed .exe loads v. slowly.
- New file can only be created as the first action, not after an existing file has been loaded.
- Data file name must end in .xlsx, but not enforced.
- No way to un-add incorrectly added group other than editing .xlsx file directly.
- Having to resize the feedback panel every time.
- Distributed .exe is large.
In the order they occur from left to right:
image_time: The time the image was captured according to the trail camera. YYYY/MM/DD HH:MM:SS. Not unique when there's more than one group seen in a single image or two different trail cameras coincidentally capture images in the same second.
group_number: Very rarely two groups of resource users may be seen in the same
image. In these cases there will be multiple records (rows) in the spreadsheet
referring to the same image (added using the Add group button). The typical case is
that there is a single record for each image, and the group_number is 1. This is true
even for images with no relevant content, e.g. triggered by animals or wind. When the
data entry person adds a second record for the same image with the Add group button,
that record will have group_number = 2, and so on if there are additional groups.
adults_n: Number of adults seen in the image (in the group the record describes, typically only one group per image).
children_n: Number of children seen in the image / group.
pets_n: Number of pets seen in the image / group.
direction: Direction of travel.
activity: Activity in which the group is engaged.
other: When activity = "other", describe here.
camera_deploy: Text identifying the camera and the deployment period, e.g. "main-gate-2022-spring" as the camera's precise position may vary between deployments.
temperature_f: The temperature in degrees Fahrenheit as shown on the image. This information is not available in the image's metadata (exif-data) and must be entered manually.
group_id: Not to be confused with group_number, this unique value identifies the
group being described by the record, and will be shared between all records in which the
same group is seen. This is the field controlled by the "Related observations" Add
and Unrelate buttons.
entry_by: The name / initials of the person doing initial data entry. Each person should always use exactly the same name / initials, the drop down menu provides values previously used.
qa_by: The name / initials of the person doing quality assurance (QA). Each person should always use exactly the same name / initials, the drop down menu provides values previously used.
observation_status: The (workflow) status of the observation. This starts off as blank or "not classified" when a record is first created for an image, and can be advanced through various states such as "entry complete" and "QA complete".
image_name: The filename of the image without path, usually something like
YYYYMMDD_HHMMSS.jpg.
image_path: The path and name of the image within the project's top level image
folder, something like site3/north_camera/2021/20210804_114454.jpg.
image_path_full: The full path to the image the last time the data was saved. This
will include the location of the project's not level image folder, which can be changed
as needed. Something like
c:\User\Jo\projects\monitoring\images\site3\north_camera\2021\20210804_114454.jpg.
image_w: The width of the image in pixels as read when the observation is created.
image_h: The height of the image in pixels as read when the observation is created.
image_bytes: The size of the image in bytes as read when the observation is created.
observation_id: A unique ID for this observation (row).
image_hash: A unique fingerprint for the image. If the image_path field was lost
or the images were re-arranged in the folder so they no longer match the image_path
field, this fingerprint could be used to re-link the images and the observation records,
as long as the images have not been altered in any way (annotated, resized, etc.).
There are four copy / paste buttons below the image:
Copy - this copies the current observation's data to the clipboard.
Paste - this pastes the data previously copied to the clipboard with the
Copy button into the current observation.
Paste+ - like Paste, but also advances to next observation.
Paste Prev. - like paste, but copies data from the immediately previous
observation, not the clipboard.
- Load both workbooks.
- Copy all of one.
- In the other, rename its one sheet (tabs at the bottom) something like "JJJ"
- Add a sheet, name it something like "AS", and paste the copied content in the new sheet.
- Close the unaltered workbook.
- Add a third sheet, and put this formula in cell A1
=IF(JJJ!A1=AS!A1, JJJ!A1, JJJ!A1 & " : " & AS!A1 & ": ?")
which reads "if content in sheet JJJ cell A1 equals content in sheet AS cell A1 show content from sheet JJJ cell A1 otherwise show 'X : Y: ?' where X and Y are the respective content from JJJ and AS".
Then to make the non-matching entries pink, select the entire sheet, then from the Home menu in the Styles "group" select the Conditional Formatting icon:
- Highlight Cells Rules
- Text that Contains...
- Then add a rule for text that contains ": ?" (colon space question-mark without the quotes) and pick a format, you can choose Custom format to make the pink background or just use one of the presets.
Now everything that doesn't match should be highlighted.
group_id, observation_id, image_full_path, entry_by, qa_by - these are not
expected to match and their columns can be hidden.
image_name is intended to match and image_path will match if you arranged
image folders the same way.
image_name might not match if filenames weren't converted to
YYYYMMDD_HHMMSS format. So this is a case where that weird image_hash
column is valuable, as long as that matches you know you're looking at the same
image regardless of filename etc.
The final step in preparing the comparison spread-sheet is adjusting for additional groups, and missing days. The mechanics are just to move the data in the sheet without the group / missing day down, creating a blank line / lines. But Excel "helps" by adjusting the formulas in the comparison sheet to account for the shift. To avoid this select the entire range that needs to move down, and move it down with the mouse while holding Ctrl. This fills in the space with copies of the top row of the moved range, those need to be selected and cleared (Delete).
python -m venv venv_tract
. venv_tract/bin/activate
pip install -r requirements.txt
# see pre-commit
mkdir -p .git/hooks; cp pre-commit .git/hooks; chmod +x .git/hooks/pre-commitconda create -n pyexe --override-channels -c conda-forge 'python>=3.9'
pip install auto-py-to-exe
start /c/Users/UserName/pkg/miniconda/envs/pyexe/Scripts/auto-py-to-exe.exe