All articles tagged with #video reconstruction
Researchers used calcium imaging of individual neurons in the mouse visual cortex and a dynamic encoding model to reconstruct 10-second videos that mice watched, with accuracy improving as more neurons were included; the method even decoded unseen scenes, shedding light on how brain representations differ from reality and enabling cross-species comparisons.
Researchers used single‑cell recordings from the mouse visual cortex to recreate 10‑second videos the mice watched, by applying a dynamic neural encoding model and comparing predicted neuron activity to actual activity. The reconstructions, which improved when more neurons were included and were validated by pixel-level correlations, offer a closer look at how visual information is encoded in the brain and why neural representations can diverge from reality, with work published in eLife.
CEBRA, a machine-learning algorithm, can compress time series data to reveal hidden structures in behavioral and neural data. It can also decode activity from the visual cortex of the mouse brain to reconstruct a viewed video. This method can be used to study brain dynamics and has potential applications in neuroscience research.