Keisuke Sehara 987c4ca6ad add performance profiling again | 3 anos atrás | |
---|---|---|
.. | ||
F01_trace-comparison | 3 anos atrás | |
F02_densities | 3 anos atrás | |
F03_conditional-probability | 3 anos atrás | |
F04_summary | 3 anos atrás | |
01_data-formatting.ipynb | 3 anos atrás | |
02_summary.ipynb | 3 anos atrás | |
README.md | 3 anos atrás | |
analyzed-data.h5 | 3 anos atrás | |
stats.md | 3 anos atrás |
For each behavioral session with real-time trigger generation, we created a single DeepLabCut project dedicated for post hoc estimation of whisker tips to use it as the "ground-truth" data.
Based on this ground-truth dataset, we computed the conditional probability of trigger generation for the whiskers being located at each position to characterize the accuracy of trigger output signals.
The analyzed-data.h5
HDF5 file contains infomation on each video.
The code to generate this dataset is found in 01_data-formatting.ipynb
, and the panels generated during the procedures are in:
F01_trace-comparison
: comparison of traces between real-time and post-hoc estimation.F02_densities
: plots of dwell-time histograms used in the paper.F03_conditional-probability
: plots of conditional probability densities used in the paper.analyzed-data.h5
Each video comprises one entry under the root, numbered as 001, 002, etc.
Attributes of each entry contains the information of the video, such as:
subject
: name of the animal.session
: name of the behavioral session.run
: the index / starting time stamp of the video.expression
: the expression used to turn the positions of whisker tips into the status of trigger output.px_per_mm
: the scale information computed from the video.realtime
and posthoc
estimations.realtime
and posthoc
store the values of expression
at given time
. The trigger
data holds the output of the evaluation.kernel
: the attributes contain the information of the Gaussian kernel (in pixels) used for kernel density estimation.positions
: the list of positions (in pixels) used to estimate the dwell-time density.all
: the values of dwell-time densities, during the whole recording period of the video, based on the post hoc estimation.triggered
: the values of dwell-time densities, when the real-time trigger was active, based on the post hoc estimation.positions
: the list of positions (in pixels) used to compute conditional probability.probability
: the values of conditional probability at given position
.sigmoid
: the attributes (in pixels) contain the information on the sigmoid curve (we used the cumulative of Gaussian) fitted to the trace of conditional probability distribution.Several types of summary data are found here:
02_summary.ipynb
is the Jupyter notebook used to summarize the data in analyzed-data.h5
.F04_summary
contains the panels generated by the above notebook (and used in the paper).stats.md
stores the statistical information analyzed in the above notebook.datacite.yml | |
---|---|
Title | Data for Sehara et al., the real-time DeepLabCut project |
Authors |
Sehara,Keisuke;Institut für Biologie, Humboldt Universität zu Berlin, Berlin, 10117 Germany.;ORCID:0000-0003-4368-8143
Zimmer-Harwood,Paul;Department of Physiology, Anatomy and Genetics, University of Oxford, Oxford OX1 3PT, United Kingdom. Colomb,Julien;Institut für Biologie, Humboldt Universität zu Berlin, Berlin, 10117 Germany.;ORCID: 0000-0002-3127-5520 Larkum,Matthew E.;Institut für Biologie, Humboldt Universität zu Berlin, Berlin, 10117 Germany.;ORCID:0000-0002-6627-0199 Sachdev,Robert N.S.;Institut für Biologie, Humboldt Universität zu Berlin, Berlin, 10117 Germany.;ORCID:0000-0002-3127-5520 |
Description | Computer vision approaches have made significant inroads into offline tracking of behavior and estimating animal poses. In particular, because of their versatility, deep-learning approaches have been gaining attention in behavioral tracking without any markers. Here we developed an approach using DeepLabCut for real-time estimation of movement. We trained a deep neural network offline with high-speed video data of a mouse whisking, then transferred the trained network to work with the same mouse, whisking in real-time. With this approach, we tracked the tips of three whiskers in an arc and converted positions into a TTL output within behavioral time scales, i.e 10.5 millisecond. With this approach it is possible to trigger output based on movement of individual whiskers, or on the distance between adjacent whiskers. Flexible closed-loop systems like the one we have deployed here can complement optogenetic approaches and can be used to directly manipulate the relationship between movement and neural activity. |
License | Creative Commons 4.0 Attribution (https://creativecommons.org/licenses/by/4.0/) |
References |
Sehara K, Zimmer-Harwood P, Larkum ME, Sachdev RNS (2021) Real-time closed-loop feedback in behavioral time scales using DeepLabCut. [] (IsSupplementTo)
|
Funding |
EU, EU.670118
EU, EU.327654276 EU, EU.720270 EU, EU.785907 EU, EU.945539 DFG, DFG.250048060 DFG, DFG.246731133 DFG, DFG.267823436 |
Keywords |
Neuroscience
Behavioral tracking Closed-loop experiment system |
Resource Type |
Dataset |