IMGStore houses your video frames
IMGStore - Houses Your Video And Data
Imgstore is a container for video frames and metadata. It allows efficient storage and seeking through recordings from hours to weeks in duration. It supports compressed and uncompressed formats.
Imgstore allows reading (and writing) videos recorded with loopbio's Motif recording system.
Video data is broken into chunks, which can be individual video files
a directory full of images
DirectoryImgStore. The format of the chunks determines if the store is
compressed, uncompressed, lossless or lossy.
Along side the video data Imgstore can also record different types of metadata:
- Recording (User) Metadata:
This includes information set at the time of recording, such as 'genotype', that is constant for the entire duration of recording.
- Frame Metadata:
This is the
frame_timestampfor every frame in the imgstore
- Extra Data:
This is DAQ and Audio data recorded by Motif at a rate different, and often faster than, the video framerate.
See Extracting Metadata.
There are only a few public API entry points exposed (most operations are
ImgStore objects (see writing and reading examples below).
open a store for reading.
new_for_format(format, path, **kwargs)
- Open a store for writing
- You also need to pass
imgshapeis the array shape, i.e.
list supports formats (remember to test after install).
extract a single frame at given frame_index from file, for example a quick way to get the 1st frame.
Images in an imgstore are identified or located 3 ways
an integer, usually directly from the camera, in the range of
N ... M. The first frame in an imagestore is not necessarily
a floating point number in seconds, in 99% of cases and camera configurations, this is the unix epoch - the number of seconds since 1970-1-1 (in UTC).
an integer, in the range
0 ... P. The first frame in the store is always
Example: Read a store
from imgstore import new_for_filename store = new_for_filename('mystore/metadata.yaml') print('frames in store:', store.frame_count) print('min frame number:', store.frame_min) print('max frame number:', store.frame_max) # read first frame img, (frame_number, frame_timestamp) = store.get_next_image() print('framenumber:', frame_number, 'timestamp:', frame_timestamp) # read last frame img, (frame_number, frame_timestamp) = store.get_image(store.frame_max) print('framenumber:', frame_number, 'timestamp:', frame_timestamp)
read images sequentially.
get_image(frame_number, exact_only=True, frame_index=None)
read an image by its
return the image nearest to the supplied timestamp.
converts the frame timestamp to a datetime with timezone information according to the store metadata which contains the timezone it was recorded in.
Example: Write a store
import imgstore import numpy as np import cv2 import time height = width = 500 blank_image = np.zeros((height,width,3), np.uint8) store = imgstore.new_for_format('npy', # numpy format (uncompressed raw image frames) mode='w', basedir='mystore', imgshape=blank_image.shape, imgdtype=blank_image.dtype, chunksize=1000) # 1000 files per chunk (directory) for i in range(40): img = blank_image.copy() cv2.putText(img,str(i),(0,300), cv2.FONT_HERSHEY_SIMPLEX, 4, 255) store.add_image(img, i, time.time()) store.close()
You can also add additional (JSON serialable) data at any time, and this will be stored
with a reference to the current
frame_number so that it can be retrieved
and easily combined later.
add_image(img, frame_number, frame_time)
adds an image to the store.
closes the store after writing, also supports the context manager protocol.
Extracting frames: frame index vs frame number
Stores maintain two separate and distinct concepts, 'frame number', which is any integer value associated with a single frame, and 'frame index', which is numbered from 0 to the number of frames in the store. This difference is visible in the API with
class ImgStore def get_image(self, frame_number, exact_only=True, frame_index=None): pass
where 'frame index' OR 'frame number' can be passed.
To get the Recording (user) metadata access the
To get all the image metadata at once you can call
which will return a dictionary containing all
To retrieve a pandas DataFrame of all extra data and associated
Command line tools
Some simple tools for creating, converting and viewing imgstores are provided
- view an imgstore
- shows frame number/time.
- 'Space' to pause/play, 'Esc' or 'q' to quit.
imgstore-save --format 'avc1/mp4' --source /path/to/input.mp4 /path/to/store/to/save
--sourceif omitted will be the first webcam
- run extensive tests to check opencv build has mp4 support and trustworthy encoding/decoding
IMGStore depends on reliable OpenCV builds, and built with mp4/h264 support for writing mp4s.
You can install opencv from pip or conda if you wish, or on linux you can use the system apt-get
installed opencv if you create a virtual environment with
Once you have a python (virtual) environment with a recent and reliable OpenCV build, you can install IMGStore from pip
$ pip install imgstore
After installing imgstore from any location, you should check it's tests pass to guarantee that you have a trustworthy OpenCV version.
If you must install opencv from pip, such as for only reading imgstores, you can use the following command to install the latest 4.8 version
$ pip install "opencv-python < 4.9"
The last version to support python2 was imgstore 2.9. (or commit 5a95d1d4454fb2a74f5a2b0d6ee7b75647b0aedb). If you need python2 support, you should install from these version/s
Post install testing
Test libraries are not installed by default. To do so, install pytest
$ pip install "pytest < 8"
You should always run the command
imgstore-test after installing imgstore. If your
environment is working correctly you should see a lot of text printed, followed by the
==== 66 passed, ..... ======
To test against the package without installing first, run
python -m pytest
Note: by running pytest through it's python module interface, the interpreter adds
PYTHONPATH, as opposed to running tests through
py.test which doesn't.
Note: if you recieve many failed tests with the error message 'The opencv backend does not actually have write support' or 'Your opencv build does support writing this codec', this is not an imgestore bug - it is a warning that you have an OpenCV version that does not support Writing h264 encoded videos. This is often the case on windows.
Even if some write tests fail due to these issues, you can stil use the imgstore package to read h264 encoded video files.
Release history Release notifications | RSS feed
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Hashes for imgstore-0.3.7-py2.py3-none-any.whl