Skip to main content

Web-Based PWA for the Visual Inspection of Cocoa Beans

We developed a web-based PWA for the visual inspection of cocoa beans. The solution combines image classification, object detection, and semantic segmentation with a practical infrastructure for data pipelines, experiment tracking, and edge devices.

Client/Company/Industry

QVISIONS GmbH

Duration

36 months

Product

Service

Expertise

Software Development

Goal

The goal of the project was to develop a web-based PWA for the visual inspection of cocoa beans. Image data was to be evaluated automatically, and the solution was designed to be used reliably both in model development and in a production-oriented environment.

Tasks

  • Implementing object detection with YOLO
  • Implementing semantic segmentation with Detectron
  • Implementing image classification with TensorFlow and Keras
  • Developing backend and frontend components for the PWA
  • Containerizing the solution with Docker and Docker-Compose
  • Designing and implementing versioned data pipelines with DVC
  • Automating experiment tracking with MLFlow
  • Setting up CI pipelines in GitLab CI/CD based on Docker containers
  • Controlling cameras via NVIDIA Jetson
  • Planning and constructing a photo box
  • Using LabelStudio for data preparation
  • Using OpenCV, OpenVINO, and TensorRT for image processing and model deployment
  • Using PostgreSQL for data storage

Challenges

A key challenge was combining different computer vision methods such as object detection, segmentation, and classification within one unified application. In addition, the data basis was difficult because assessments of the cocoa beans were sometimes contradictory. As a result, data collection, model training, and evaluation had to be aligned very carefully to achieve reliable results.

Programming Languages

Python

Technologies

DVC, Detectron, Docker, Docker-Compose, Flask, GitLab CI/CD, Keras, LabelStudio, MLFlow, NVIDIA Jetson, OpenCV, OpenVINO, PostgreSQL, PWA, TensorFlow, TensorRT, YOLO

Project Image

Web-based application for the visual inspection of cocoa beans with camera control, image analysis, and AI-supported evaluation.

Similar problem?

Contact us

Takeaway

The result was a web-based solution for the visual inspection of cocoa beans with integrated computer vision and machine learning components. Versioned data pipelines, experiment tracking, and the integration of edge devices created a solid foundation for practical use.

Similar Projects

Project Image

RIM2D - Highly Efficient 2D Hydraulic Simulation of Fluvial, Pluvial, and Urban Flooding

Hydrodynamic Simulation Web Application Geodata GPU Computing

RIM2D is an existing, highly efficient 2D hydraulic simulation model for fluvial, pluvial, and urban flooding. As part of a strategic partnership, we supported the extension of the research code with a web application and a cloud-based GPU simulation environment, enabling its transition into a market-ready product.

Project Image

Data Lake for Geoscience Research Data Management

Datalake Research Data Management Geodata Cloud-Native Open-Source

We developed an Open-Source S3-based data lake solution for the centralized ingestion, categorization, and searchability of data. The goal was to automate and improve manual data management through an integrated architecture with workflow orchestration, data cataloging, and access control.

Back To Top