SynthDet

SynthDet is a synthetic image dataset created for training object detection models in retail environments. It includes 3,500 images split into training (2,450), validation (175), and testing (875) sets. All images are generated within a virtual supermarket environment, ensuring full control over camera angles, lighting, and product placement. SynthDet is tailored to match the deployment scenario: cameras positioned perpendicularly to the shelves for a clear frontal view. This setup ensures that the synthetic data mimics real deployment conditions while maintaining variability in product layout, lighting, and shelf configuration. By narrowing the domain gap between training and deployment environments, SynthDet enhances model generalization and accuracy in real-world shelf detection tasks.