Solving Logistic Problems with Computer Vision: a Case Study

How an alwaysAI user is using computer vision with IoT to solve logistics problems.

As always, we are proud of our users and what they are accomplishing by using alwaysAI to integrate Computer Vision into their projects. This week we would like to highlight our user Abhijeet Bhatikar who is using alwaysAI as part of his project for the OpenCV Spatial AI Competition. Read on to find out more about how he is combining alwaysAI, computer vision and depth cameras to help solve logistics problems. 

Sign-up here to try alwaysAI!

Abhijeet is an IoT developer at The Dock in Dublin Ireland which is Accenture’s flagship R&D and global innovation center. The Dock is a diverse team of over 250 cross-industry problem-solvers bringing design, business, and technology together to help solve some of the most complex problems that their clients are facing today. Abhijeet leads the IoT team for Rapid Innovation, which is a practice within the Dock. He is currently participating in the OpenCV Spatial AI Competition which is sponsored by Intel. The purpose of this competition is to get developers to build applications with a smart camera that combines the power of neural inference with depth perception in real-time.

Abhijeet’s project for the competition is backed by Accenture, meaning they give him the time to work on it. However, he would still want to participate without company support, as he finds computer vision and AI coupled with IoT fascinating. The project that his team is working on for the OpenCV Competition is all about maximizing freight carriers' warehouse space.

Generally, when it comes to storing packages in a warehouse, there is no particular method involved in optimizing space. Warehouse space comes at a cost. As per the research, there is about a 15%-25% void in package warehousing efficiencies. Abhijeet’s team’s idea is to use computer vision on depth cameras to optimally organize packages, maximizing space usage by detecting packages’ shape, dimensions and weight. He is using the OAK-D camera to do neural network inferencing as well as calculate the depth at the edge. The OAK-D has an Intel Movidius Myriad X image processor, meaning it can run image processing at the edge pretty well. In Europe, this is key as most computer vision application projects don’t go to production because of privacy issues and regulations around data collection and data storage. By running computer vision applications at the edge, data is neither sent nor stored in the cloud, which is ideal from the compliance perspective.

Abhijeet is an IoT developer and did not have extensive AI and computer vision experience prior to using alwaysAI. Learning to build computer vision applications usually entails a pretty steep learning curve.

Prior to the OpenCV competition, he had tried out a few personal computer vision projects using OpenCV. He used Azure IoT Edge to run them on the edge. When he discovered alwaysAI, Abhijeet found that it augments the developer’s ability even further, enabling computer vision and AI applications to be built at the edge faster. Installing OpenCV on a Raspberry Pi by building from source takes a very long time. As a developer, it's a loss of time that could be better spent working on the actual problem. It is essentially a barrier to entry into the computer vision app development space, especially if you don’t want to use the cloud for privacy and cost purposes. There are several benefits of running computer vision on the edge, however, doing it without a platform like alwaysAI would take a lot of time to prototype.

For Abhijeet, alwaysAI cuts the development time down substantially. When comparing alwaysAI to other similar platforms, Abhijeet believes alwaysAI combines the best of all worlds; running computer vision models on the edge, fewer privacy concerns, the availability of pre-trained models, the ability to customize models, cost-effectiveness and ease of use. The support for the developers from the alwaysAI team has also been of great help. For him, the biggest benefit was the time he saved using alwaysAI. Being able to get up and running and prototyping quickly also means going into production faster. 

When Abhijeet first started using alwaysAI, he saw the platform would work differently on different platforms; either as a docker container or as a native app on virtual environments. He had experience with Docker containers before and wanted to run his apps that way. He said that he had a lot of questions on running his computer vision app as a docker container, and Eric, one of our Senior Software Engineers, was helpful and answered all of his queries on Discord. He also engages with other users on Discord and helps out, answering their questions wherever he can. He is one of our most engaged users, and we are certainly grateful for all his feedback and participation in our webinars and Hacky Hours! 

The OpenCV competition concludes at the end of October, so by this time, he will have a working prototype with the OAK-D camera. We look forward to hearing how his team does and will be sure to update our community! 

Disclaimer: All of the opinions expressed in the article are personal opinions of the user and do not represent the company he works for. Abhijeet is speaking on his behalf and not on behalf of his team or Accenture.