Have a look around our new website for the discovery and sharing of research data and let us know what you think. See How to Submit for instructions on how to publish your research data and code.
Pereira, Talmo D.; Aldarondo, Diego E.; Willmore, Lindsay; Kislin, Mikhail; Wang, Samuel S.-H.; Murthy, Mala; Shaevitz, Joshua W.
Abstract:
Recent work quantifying postural dynamics has attempted to define the repertoire of behaviors performed by an animal. However, a major drawback to these techniques has been their reliance on dimensionality reduction of images which destroys information about which parts of the body are used in each behavior. To address this issue, we introduce a deep learning-based method for pose estimation, LEAP (LEAP Estimates Animal Pose). LEAP automatically predicts the positions of animal body parts using a deep convolutional neural network with as little as 10 frames of labeled data for training. This framework consists of a graphical interface for interactive labeling of body parts and software for training the network and fast prediction on new data (1 hr to train, 185 Hz predictions). We validate LEAP using videos of freely behaving fruit flies (Drosophila melanogaster) and track 32 distinct points on the body to fully describe the pose of the head, body, wings, and legs with an error rate of <3% of the animal's body length. We recapitulate a number of reported findings on insect gait dynamics and show LEAP's applicability as the first step in unsupervised behavioral classification. Finally, we extend the method to more challenging imaging situations (pairs of flies moving on a mesh-like background) and movies from freely moving mice (Mus musculus) where we track the full conformation of the head, body, and limbs.
Amazonian deforestation causes systematic changes in regional dry season precipitation. Some of these changes at contemporary large scales (a few hundreds of kilometers) of deforestation have been associated with a ‘dynamical mesoscale circulation’, induced by the replacement of rough forest with smooth pasture. In terms of decadal averages, this dynamical mechanism yields increased precipitation in downwind regions and decreased precipitation in upwind regions of deforested areas. Daily, seasonal, and interannual variations in this phenomenon may exist, but have not yet been identified or explained. This study uses observations and numerical simulations to develop relationships between the dynamical mechanism and the local- and continental-scale atmospheric conditions across a range of time scales. It is found that the strength of the dynamical mechanism is primarily controlled by the regional-scale thermal and dynamical conditions of the boundary layer, and not by the continental- and global-scale atmospheric state. Lifting condensation level and wind speed within the boundary layer have large and positive correlations with the strength of the dynamical mechanism. The strength of these relationships depends on time scale and is strongest over the seasonal cycle. Overall, the dynamical mechanism is found to be strongest during times when the atmosphere is relatively stable. Hence, for contemporary large scales of deforestation this phenomenon is found to be the prevalent convective triggering mechanism during the dry and parts of transition seasons (especially during the dry-to-wet transition), significantly affecting the hydroclimate during this period.
Battaglia, D. J.; Boyer, M. D.; Gerhardt, S.; Mueller, D.; Myers, C. E.; Guttenfelder, W.; Menard, J. E.; Sabbagh, S. A.; Scotti, F.; Bedoya, F.; Bell, R. E.; Berkery, J. W.; Diallo, A.; Ferraro, N.; Jaworski, M. A.; Kaye, S. M.; LeBlanc, B. P.; Ono, M.; Park, J. -K.; Podesta, M.; Raman, R.; Soukhanovskii, V.
