Unveiling the Invisible: How Your Phone's Camera Sees More Than You Think

Nikhil Behari, a graduate student at MIT's Camera Culture group and NASA fellow, invites us to reconsider the potential of everyday smartphone photos. Beyond mere snapshots, Behari argues that these images capture hidden dimensions of our environment through shadows, reflections, and even individual photons, offering a wealth of information that AI systems are only beginning to decipher.

The Hidden World in Plain Sight

Behari's journey begins with a familiar scene: exploring the campuses of MIT and Harvard. These everyday photos, however, hold more than meets the eye. Behari directs our attention to shadows, ubiquitous yet often overlooked elements that reveal a surprising amount about our surroundings. From the length and angle of shadows, we can infer the time of day, the height of objects, and even the presence of individuals not directly visible in the frame. Shadows, in essence, act as a hidden camera, capturing aspects of a scene that would otherwise remain unseen.

Decoding Shadows: A Challenge for AI

While humans intuitively understand the information conveyed by shadows, AI systems struggle to interpret these subtle cues. Behari's research at MIT focuses on bridging this gap, designing AI capable of extracting meaningful insights from shadows. One application of this technology is the creation of city-scale 3D models from satellite imagery, using shadows to reconstruct the urban landscape.

Unveiling the Invisible: How Your Phone's Camera Sees More Than You Think
The hidden information in your photos | Nikhil Behari | TEDxMIT

Reflections: Mirrors to Another World

Reflections offer another layer of hidden information within our photos. Behari uses examples ranging from building reflections to the reflections in a car mirror, to illustrate how reflections reveal aspects of our surroundings that would otherwise be hidden. AI systems face challenges in interpreting reflections due to the warping of the 3D world on 2D surfaces and the mixing of colors between the reflected scene and the reflecting object.

Behari's work addresses these challenges by developing AI algorithms that can disentangle the color and reflection components of an image, effectively turning any reflective surface into a window onto the surrounding environment.

The Power of Photons: Seeing the Unseen

Delving deeper into the capabilities of smartphone cameras, Behari illuminates the role of LiDAR (Light Detection and Ranging) technology. This technology emits hundreds of lasers, measuring the time it takes for photons to return to the camera. By analyzing the time delay of these returning photons, smartphones can create depth maps of the surrounding environment, enabling features like autofocus.

Behari's research takes this a step further, visualizing individual photons in flight. By capturing the movement of photons in ultra-slow motion, his team can uncover hidden information about the way light interacts with objects, revealing details about an object's geometry and material properties. This technology has the potential to unveil objects that are hidden from direct view, opening new possibilities for AI-powered vision.

Implications and the Future of AI Vision

Behari's work has significant implications for the field of artificial intelligence, demonstrating the potential to extract far more information from everyday images than previously thought. By training AI to interpret shadows, reflections, and individual photons, we can unlock new capabilities in areas such as 3D modeling, object recognition, and scene understanding.

Conclusion: A New Perspective on Photography

Behari concludes by encouraging us to explore our own camera rolls with a fresh perspective. By looking closely at the shadows, reflections, and subtle details within our photos, we can begin to appreciate the hidden information captured by our smartphones. As AI technology continues to advance, the potential to unlock even more insights from these everyday images is limitless, paving the way for a future where our cameras can truly see the invisible.

3 min read