For any given species, the design of an animal’s visual system reflects the challenges of its ecological niche; thus, a promising approach to study visual system function is to probe the system with natural stimuli. Mice have become an important model in vision research, but it is still rarely considered that, compared to primates, they live in a different environment and therefore have different visual needs. For example, unlike primates, mice are dichromatic and perceive UV light. Moreover, the mouse retina is subdivided into a mostly “green” sensitive (peak at 510 nm) dorsal and UV sensitive (peak at 360 nm) ventral retina. Therefore, presenting naturalistic stimuli in laboratory settings to non-primate species, such as mice, is challenging.Under the assumption that a substantial fraction of mouse eye movements serves to stabilize the retinal image, we built a gimbal-stabilized, spectrally-calibrated hand-held camera to explore the natural habitat of mice in the relevant spectral bands. We intensity-calibrated the camera with LEDs of defined wavelengths and brightness using a power meter / spectrometer combination. The camera was moved close to the ground along mouse tracks and UV/green movies of the mouse habitat were recorded for different representative scenes and at different times of the day. By analysing contrast statistics of the movies, we found, for example, that contrast in the two chromatic channels (UV/ green) diverged greatly in the upper but not in the lower visual field. This resonates well with reports of a higher fraction of colour-opponent retinal ganglion cells in the ventral mouse retina and superior behavioural colour discrimination in the upper visual field. In addition, we found that during dusk and dawn, “predators” coming from the sky should be more easily detectable in the UV compared to the green channel, which emphasizes the UV’s role for mouse vision. Finally, we designed different unsupervised models, and when fitting them to our recordings, we mainly found color-opponent filters with training data of the upper visual field. In the last part of the talk, I will also show ongoing efforts to established a light-weight, head-mounted camera system, which can capture the visual environment from the perspective of freely roaming mice.