Source: VIMAAN ROBOTICS, INC. submitted to NRP
AUTONOMOUS AI BASED MONITORING AND IMPROVEMENT OF PACKAGING AND LABELING FOR MEAT PROCESSING
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
ACTIVE
Funding Source
OTHER GRANTS
Reporting Frequency
Annual
Accession No.
1029908
Grant No.
2023-70439-39186
Cumulative Award Amt.
$1,000,000.00
Proposal No.
2023-01076
Multistate No.
(N/A)
Project Start Date
Feb 15, 2023
Project End Date
Aug 14, 2025
Grant Year
2023
Program Code
[MPPRI]- Meat and Poultry Processing Research and Innovation
Recipient Organization
VIMAAN ROBOTICS, INC.
3180 DE LA CRUZ BLVD # 280
SANTA CLARA,CA 950542434
Performing Department
(N/A)
Non Technical Summary
Vimaan Robotics, Inc. (Vimaan) is pleased to respond to this RFA for NIFA's Meat and Poultry Processing Research and Innovation (MPPRI). Vimaan is a small business entity for process and workflow automation, using Artificial Intelligence and Machine Learning for monitoring, tracking and tracing of food, driving lower food cost, improved safety and higher quality for diverse community needs. Vimaan's proposal aims to address vulnerabilities that Small to Medium-Sized Meat and Poultry Processors (SMMPPs) face from larger meat and poultry processors with similar automation technologies.Vimaan will achieve this with minimal additional research to improve the supply chain resilience of SMMPPs. Vimaan will work with two SMMPPs in underserved communities, addressing this RFA's Priority Area 1: "Monitoring and Improving Complex Processes", aligning partially or fully with USDA Strategic Goals 2 to 5 for FY 2022-2026. Vimaan will demonstrate a sound approach to generate higher revenues and profitability for SMMPPs, reducing waste at scale, with field validation and commercialization of this innovation.Vimaan will use its proprietary machine learning to read labels on meat packaging and also examine the seal integrity and quality on these packages to ensure that the quality is maintained at a high level. This will help reduce food related contaminations, food wastage due to returns and destruction, and overall, allow small and medium meat processors to become more efficient and profitable.Vimaan has an active NSF SBIR Phase 2 award to use robots, computer vision and AI to enhance inventory visibility, traceability and product quality in warehouses. Vimaan's technology is commercially live at several Fortune 500 companies, including Food & Beverage distribution, where Vimaan's customers use the technology to highlight expiry date issues, damaged packaging, and incorrect order fulfillment (which can result in food condemnation).
Animal Health Component
(N/A)
Research Effort Categories
Basic
(N/A)
Applied
(N/A)
Developmental
100%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
51233202080100%
Knowledge Area
512 - Quality Maintenance in Storing and Marketing Non-Food Products;

Subject Of Investigation
3320 - Meat, beef cattle;

Field Of Science
2080 - Mathematics and computer sciences;
Goals / Objectives
The meat processing industry faces common quality problems in product packaging and labeling, which can lead to health safety issues, regulatory non-compliance, and meat recalls. SMMPPs expend considerable expense on workers to inspect each package. But this approach is expensive, and besides, COVID has exposed challenges both with worker proximity as well as availability. A fully automated solution would therefore greatly improve SMMPPs capabilities.Vimaan proposes to use its autonomous Computer Vision and Artificial Intelligence platform to solve this problem. Vimaan will implement its technology at the packaging and labeling process steps during processing to enable 100% real-time automated inspection of each package for compliance. This will highlight quality issues as they occur, which will in turn enable SMMPPs to reduce waste, improve product tracking and traceability, and ultimately improve quality, and safety. Vimaan's autonomous solution for monitoring, tracking and tracing would help SMMPPs build resiliency over the longer-term, as shown by larger processors.Vimaan will work with its two partner SMMPPs to conduct a pilot project in an actual meat processing environment to demonstrate the technology capability at scale. Metrics to assess the solution's efficacy will be identified and tracked during the pilot project. Using these metrics, the performance of the production line using Vimaan's technology will be compared against existing manual quality processes currently in place at the processor.The pilot project will allow Vimaan to reach a TRL 9. As part of the project, Vimaan will also pursue broader commercialization of the technology across the domestic SMMPP sector by conducting ROI and techno-economic analyses to identify the value proposition, generate a go-to-market plan and initiate business development activities with various target SMMPPs.
Project Methods
Solution Overview: DockTRACK Parcel and PackVIEWVimaan Robotics ("Vimaan" or "Company") is responding to USDA's Meat and Poultry Processing Research and Innovation (MPPRI) Funding Opportunity Number USDA-NIFA-OP-009506, which aims to maintain and improve food supply chain resiliency. Vimaan is targeting Priority Area 1: Monitoring and Improving Complex Processes. Vimaan proposes a pilot project using its innovative automated Computer Vision, Machine Learning and Artificial Intelligence (AI) technologies to enable meat package monitoring, improve product tracking and traceability, and ultimately improve quality, safety and resiliency in the industry (the "Pilot Project"). Vimaan will implement the Pilot Project at two small meat processors to address representative quality issues that are widely experienced across the meat processing industry.Vimaan has received an NSF SBIR Phase 2 award for the use of robots and AI to capture, track and validate inventory in warehouses. During the Phase 2, Vimaan conducted extensive work using computer vision and AI to enhance inventory visibility, traceability and product quality. Vimaan uses sophisticated cameras and sensors to capture data from a work environment and processes this data through a complex AI software pipeline to read labels, measure dimensions, capture box contents, identify anomalies, etc. The aim of this application is to monitor package quality, validate labeling and flag quality issues.Vimaan will use its previously developed technologies for other applications to accomplish the goals for this project.DockTRACK Parcel[1] is a computer vision station - mounted on conveyors or table-tops - that scans packages and labels with high speed and fidelity. DockTRACK Parcel captures 1D/2D barcodes, reads label text, inspects packages for defects, dimensions the package, and reports anomalies against prior database entries. DockTRACK Parcel includes ViewDECK, Vimaan's web application, which is an intuitive tool that supports package searches, reporting and includes a comprehensive photographic historical archive of all captured packages. Figure 1 shows the conveyor belt version (DockTRACK Parcel) and the table-top version, called PackVIEW[2].[1] https://vimaan.ai/resources/video/parcel-receiving-and-shipping-with-docktrac/[2] https://vimaan.ai/resources/video/order-packing-with-100-accuracy/

Progress 02/15/23 to 02/14/24

Outputs
Target Audience:Vimaan's target audience for this reporting period included Small and Medium Meat Processors (SMMPs) who are interested in using Computer Vision and Artificial Intelligence to improve the quality of their packaged meat products, and to quickly identify and eliminate problems associated with improper packaging, improper labeling, improper order fulfilment etc. in their meat processing lines. The solution would be implemented in an automated manner that is consistent with high volume processing of packaged meats -- thus enabling high quality andcompliance with USDA regulatory requirements, without incurring high labor costs. Changes/Problems:There were two key changes within the scope of this project, mainly for the practicality of the value to the partner, and also for the phased rollout of the solution. The first change involves the inclusion of other hardware in the facility intended to address workflow errors that were beyond just during the packaging step. (During the packaging step, meat sticks are moving down a conveyor line, and automatically get packaged in large batches.) Vimaan identified two risks to directly addressing the problem at this workflow step. First, the meat was "moving" down the conveyor line, and would need to be inspected during this motion. Second, there are multiple meat sticks coming off the line concurrently; therefore, any AI solution would need to be able to process camera outputs in parallel. It was Vimaan's assessment that together, these two problems would pose additional risk. Therefore, as a first phase to prove the viability of the technology, Vimaan and its partner chose a secondary, downstream, workflow step to implement the solution. This step is the Quality Assurance step where only a few sticks are inspected in a "stationary" mode. Automation of this step would not only still add value for the Partner; it would also prove the capability of the AI to identify defects and highlight them. In addition, Vimaan would be able to deliver a vision solution at these QA Stations within the facility and use those to prove the software capabilities necessary for the success of the conveyor solution. Once the technology is proven feasible at this step, the next step would be to integrate it with the conveyor belt to allow "inline" quality inspection of 100% of the meat sticks as they rolled off the packaging line - all in real time. The second key change involves a clarification to the deliverable for identifying whether packaging was sealed correctly. This feature will still be developed for visual anomalies in the packaging, but as Vimaan's partner has pointed out, many of the cases of this failure are not visually identifiable and require the auditor to feel the product physically and identify if there is any compromised seal. Therefore, this use case may not be a good fit for Computer Vision/AI based solutions. What opportunities for training and professional development has the project provided?During this project, several of the Company's engineers had an opportunity to work on three different fronts in various capacities. 1. Engineers were able to get direct access to interactions with the customer / partner, and understand how to define a problem statement, propose solutions, and present potential solutions to the customer with all the metrics and criteria of interest. They also attended various meetings with the customer to understand the nuances of defining and presenting a solution to various stakeholders within a customer organization. 2. Engineers also got the opportunity to work with mechanical and system designers to understand and simulate the workflow, the user behavior and to design a system according to all of the needs of a customer and a workflow. 3. Finally, engineers got to work with senior technical personnel (CTO and PhD level scientists) to understand more about machine learning and computer vision models and how they may apply to modern day industrial and commercial problems. How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?The key high-level goals for the next reporting period are to successfully deploy the two variations of solutions (including hardware and software integration) intended to automate and streamline the existing validation processes and make it more reliable through computer vision and AI. These two solutions are the Pre-Packaging QA Station to validate dimensions and meat quality and the Post-Packaging QA station to validate weights and label fields. As a stretch goal, Vimaan and its Partner will also attempt to assess the feasibility of using the solution at the Conveyor station to validate meat and label quality. To achieve the goals, Vimaan will also need to develop the computer vision modules necessary to differentiate between "good" product and "bad" product. This filtering of product includes the capability to validate meat quality, label quality, label fields, weights and product dimensions. Throughout the first program year, Vimaan has worked to collect the information and product data necessary to create system and solution specifications for each of these documented goals. There has been a lot of progress made towards these goals providing optimism that the first goal is achievable within the second program year. The stretch goal naturally has more risk associated with it.

Impacts
What was accomplished under these goals? This project is aiming to monitor the consistency of the product passing through Small and Medium Meat and Poultry Processing Plants (SMMPP) and reduce the waste resulting from incorrect labeling or damaged product. This monitoring will not only save the SMMPPs money from reduced waste but will also serve to prevent damaged or expired goods from reaching the consumer. During the first reporting period Vimaan and its key partner in the SMMPP sector were able to define the processing workflow and understand the pain points. Based on an extensive understanding of the flow and the workflow locations where processing or human error could occur and where the SMMPP could derive benefit from automation, the two parties identified one specific workflow - the Quality Audit section - where the Vimaan automated Computer Vision and Artificial Intelligence technology could provide significant benefits. The key achievements reached were to generate samples of "good" and "bad" product - examples of the types of errors that typically occur during meat packaging, as well as examples of acceptable and USDA compliant packaging for product. Vimaan used these samples to generate catalogs of classes of errors, which were then used to train the AI models to automatically recognize defective product using its hardware and software solution. Further, to make it usable by its SMMPP partner, Vimaan used the output of the AI solution to integrate into its partner's existing daily product sheets. Vimaan also custom designed a software validation system and also developed a hardware solution fine-tuned to address the product going through the partner's processing line. One notable additional achievement in the first phase was the integration of the facility weight scales, used at the QA stations, to the Vimaan system. The integration with the weight scale will serve an important role in Vimaan's ability to validate the consistency of the product being produced and help ensure that the product stays within the defined guidelines while also enabling a more significant sample size of product to be evaluated. The key impact of this program will be to allow SMMPPs to QA their product more effectively and flag issues immediately when they occur. This will also allow the SMMPPs to capture a much more representative dataset which can help inform where there are other efficiency gaps or product quality inconsistencies.

Publications