Background 

The client, an agritech company, approached us to check the novelty of a vehicle-mounted AI-driven plant treatment system designed to detect and treat plants. The system leverages advanced image segmentation techniques and is equipped with a camera module, plant treatment device, transformer-based unit, transformer-based encoder-decoder model, probability matrix generation, neural networking components, and other key gears. 

Client’s objective 

The client required an invalidation search to identify prior art related to a novel image segmentation method for classifying plant and weed species. The method follows a specific process flow involving calculations in the encoding and decoding stages of the segmentation model. Their primary goal was to identify the exact prior art covering the segmentation method using transformer-based encoder-decoder architecture; the flow of processes within the encoding & decoding stages; and the probability matrix generation technique linked to selective plant treatment. 

Our Strategic Approach 

Our thorough search strategy involved an extensive exploration of the limited prior art within agriculture-focused AI systems. To ensure a comprehensive understanding, we broadened our scope to other industries employing similar segmentation methods by: 

1. Identifying a Relevant Non-Patent Literature (NPL) 

2. Addressing the Probability Matrix Argument 

3. Exploring Additional Reference on Multi-Species Treatment 

4. Drafting Segmentation Maps with Multiple Objects

Gain deeper insights into our comprehensive IP strategy and discover how our expertise helped our client successfully challenge the novelty of a patent, strengthening their legal position. 

Fill out the form to access the complete case study.