For enterprise retailers, verifying whether in-store products follow the planograms (i.e., shelf layout) prescribed by head office has been a challenge for many years. This process has likely involved thousands of hours of manual comparisons for countless retail businesses over the past decade.
Off the back of an exciting R&D effort, we developed a suite of image recognition models to help solve this headache once and for all. We’re excited to add image recognition to our arsenal of AI tools.
Here’s a quick demonstration of them working hard to identify products from images.
What is EIRS and what does it do?
EdgeRed Image Recognition Suite (EIRS) takes advantage of a range of specialised models to complete tasks and generate outputs that would traditionally be impossible with a single AI model. Much like an assembly line, each group of models completes their task and passes the work to the next team to refine.
Here’s how EIRS works:
Detecting components: EIRS starts by detecting various components in the image, such as products, shelves, and tags, and breaking them up into individual elements.
Identification process: These elements are then passed down for identification as actual products. We’ve used a mix of public and proprietary datasets to enhance classification accuracy.
Human review: We even built a web interface to enable humans to seamlessly review EIRS’ work when necessary.
Leveraging ChatGPT-4o
AI models always have limitations, for example, what happens when there are gaps in the shelf when things are out of stock? You can’t detect something that’s not there! To further the analogy of a team, we ‘recruit’ a new model that specifically detects gaps in a shelf and match that to the price tag to identify which product should have been there. We even brought in the newly minted ChatGPT-4o to help us speed up the training process on product classification.
There’s so much more we can nerd out to but I’ve been told to keep the blog posts to a certain length so please reach out to us if you’re interested so I can share all the things our new team of models is capable of. 🤓
Officially launching Planoverse! 🪐🚀
I am thrilled to announce the launch of Planoverse, our new software start-up aimed at taking our EIRS R&D work to the next level. With a dedicated team, Planoverse acts as a connector between physical stores and their digital counterparts, transforming images of real-world shelves into 3D assets.
Why Planoverse? Because it allows retailers to take advantage of planning, prediction, and automation features enabled by having digital twins of their stores.
Want to know more? We’ll be sharing more updates over the next few months, Make sure you subscribe to our newsletter to stay up-to-date.
So, how did the Planoverse come about?
Though team capacity is rare, our founders at EdgeRed, Mon and Howard, have always been committed to carving out time for R&D projects that bolster our internal capabilities and develop innovative solutions for key client challenges.
We completed multiple rounds of feasibility studies and developed a prototype before presenting it to clients, ensuring we could deliver on our promises. What began as an R&D initiative has now grown into a valuable AI capability that offers significant productivity improvements for our clients.
I am also excited to share that Planoverse was recently recognised as one of the "Top 10 AI companies tacking Australia’s biggest challenges" by Stone & Chalk & CSIRO’s National Artificial Intelligence Centre.
I’m incredibly grateful for the opportunity to lead this venture and can’t wait for what’s to come. Watch this space for more updates on Planoverse, or contact us for a demo on how this technology can benefit your business. 🤝✨
Frequently Asked Questions (FAQs)
How accurate is Planoverse in identifying products on shelves?
Planoverse leverages advanced image recognition models trained on a combination of public and proprietary datasets to ensure high accuracy. Our multi-step process, including human review when necessary, helps maintain reliability even in complex retail environments.
Can Planoverse handle different store layouts and product assortments?
How can Planoverse improve productivity for retailers?
This blog is written by Howjer, with assistance from Erica.
Interested in joining our team?
Applications for FY25 Graduates are now open! We are looking for Graduate Data Analysts and Graduate Data Engineers to join our team. Applications close 31st July 2024.
About EdgeRed
EdgeRed is an Australian boutique consultancy specialising in data and analytics. We draw value and insights through data science and artificial intelligence to help companies make faster and smarter decisions.
Subscribe to our newsletter to receive our latest data analysis and reports directly to your inbox.
Comments