In the start-up world, it’s important to take a moment and celebrate milestones even during long workdays and hectic schedules. Today is one of those moments, as we stop to thank and appreciate the grant and confidence boost we received from the Israel Innovation Authority.
The Israel Innovation Authority, which is also known as the Office of the Chief Scientist of Israel's Ministry of Economy, is focused on supporting and developing Israeli technology solutions and the companies creating them. In the current competitive landscape, a small percentage of the many startups applying for the grant are fortunate enough to receive it, which makes this achievement that much more important.
Our vision in 3D
As part of the application process, Hexa3D presented a detailed technological vision to the authority, explaining how our platform performs the retrieval, synthesis, and generation of 3D assets based on 2D images. The vetting process focused on the level of technical and functional innovative capabilities, the financial aspects related to the solution, and Hexa3D’s ability to execute the company’s vision and business goals. We demonstrated our path towards a major breakthrough in the field of 3D reconstruction by explaining the transformation of 2D images to 3D models using deep learning capabilities.
Hexa3D’s Reconstruction Engine is trained on a proprietary dataset containing tens of thousands of high quality textured 3D models, to automatically generate high fidelity 3D models from product images. In the video: depiction of the neural network process of learning to reconstruct the 3D model of a chair.
In the above video you can notice the way Hexa3D handles products clustering and similarity detection. This process is highly effective when working in a 3D reconstruction pipeline.
Current Photogrammetry methods impose significant constraints on the amount and nature of the input images, heavily limiting their applicability to real-world problems. Specifically, it is not feasible to impose major constraints on retailers, who handle huge amounts of products. We identify these constraints as the major barrier which has previously prevented the deployment of 3D reconstruction as a service in the retail industry. On the contrary, our reconstruction pipeline assumes little to nothing on the number and nature of the input images.
In supporting such algorithms, a vast 3D model dataset is required to improve the performance of these Artificial Intelligence (AI) models. Hexa3D has not only the technological expertise in training AI models, but due to its artist management platform, it has also achieved an unmatched ability in creating proprietary 3D model datasets.
Hexa3D’s Embedding Engine is trained on millions of products. The embedding engine understands concepts of category, complexity, geometry and texture, to automatically classify, price and retrieve similar products. In the video: a 3D visualization of the embedding space for the attribute of product category.
Nowadays, the industry is shifting towards utilizing 3D from the product conceptualization stage and throughout the lifecycle in multiple different use cases, which renders methods such as Photogrammetry irrelevant.
A message from start-up nation
For decades, grants given by the Israel Innovation Authority paved the way to the next big thing in global tech. This grant sends a strong message regarding the faith and confidence of the authority in the immersive content field in general, and Hexa3D’s solution in particular.
As the world faces countless changes due to the COVIS-19 crisis, certain technologies prove to be more helpful and crucial than we had imagined. We have every reason to believe that the speedy market education process the world is currently experiencing will push 3D and AR solutions further. With the recent backing of the Israel Innovation Authority, Hexa3D’s unique capabilities place it at a strategic market position, making the company a key player in the immersive content arena.
Want to know more?