Founders: Ugam Kamat, Siddharth Jha
Digital Twins (virtual representation of a physical system or process) as a concept has been around for the last few years and has been applied successfully in small-mid scale setups like buildings, factories, and manufacturing plants. But what if we could create digital twins of larger structures like a stadium, a fort, or even a city? And that too, in 3D. The richness of experience and analysis 3D digital twins can offer is unparalleled and this is what Preimage envisions achieving.
The underlying technique behind 3D representation - photogrammetry or 3D reconstruction converts a series of overlapping 2D images to a 3D model of real-world environments. 3D reconstruction/photogrammetry step can take up to 50% of the process time (from image capture to analysis), depending on the use case and number of images. Photogrammetry is not a new concept; in fact, it is centuries old. However, the current solutions are mostly legacy tools built on classical geometric approaches and hence have multiple challenges as follows:
- They are slow and compute heavy, with the processing time increasing exponentially depending on the number of images (ex: a set of 5,000 images could take from 1 to 2 days)
- Need a lot of manual post processing because of noise in point clouds, inaccurate classifications of the 3D scene, and artifacts in 3D meshes
- Fragmented space - customers choose different tools depending on the subject. The industry lacks a single tool that can be used across all use cases
The business impact of these issues is slower decision making, high cost, and longer time to market. For example, in the drone industry, inspection and monitoring requires speed and drone service providers can lose contracts because of outdated information generated due to long processing time.
The gap between increasing demand for 3D content and challenges with the existing tools is the opportunity that Ugam Kamat and Siddharth Jha set out to tap into through Preimage. They are building a cloud native new age photogrammetry platform that is
- >5x faster
- Has higher accuracy and hence can reduce manual correction by 2x
- Scalable across all use cases - small/big structures to landscapes
Early results from benchmark study show that Preimage is significantly faster than competitors
Preimage is able to achieve this through several innovations across the photogrammetry pipeline - synthetic and procedural data generation, adoption of 3D deep learning and neural rendering. In addition, the distributed computing architecture lowers compute overhead and gives them speed. Read more about Preimage value proposition and technology in this whitepaper.
Building this platform requires talent that is not easily available. Since the inception of the company, the founders have taken promising recent graduates and interns from colleges and groomed them to deliver this complex pipeline. Ugam and Siddharth’s tech and business prowess and extensive work in AI and Computer Vision in particular makes them best suited to innovate in this industry.
While the drone industry (mapping, monitoring, inspection across mining & construction etc) is the low hanging fruit, Preimage plans to target emerging use cases like gaming, metaverse, and AR/VR over the next few quarters. The company fits well into our thesis of backing innovative companies solving big global problems - in this case, disrupting an existing market as well as creating a new one.