Vectra Engine

Meta-Inference Capabilities

Build custom vision models
at lightning speed.

Initialize your few-shot learning session. Define the number of distinct classes you want your model to recognize, and we'll provision the environment instantly.

Data Ingestion

Upload your Support (Training) and Query (Testing) assets.

Select Architecture

Choose the foundation model. Heavier models offer more accuracy but compute slower.

Architecture Guide

  • ResNet (18, 34, 50): Reliable & robust. Deeper networks (higher numbers) yield better accuracy at the cost of processing speed.
  • MobileNet (V2, V3): Highly optimized for edge/mobile devices. Extremely lightweight.
  • EfficientNet (B0, B1): State-of-the-art scaling technique for maximum parameter efficiency.
  • DenseNet (121, 169): Excellent for complex visual patterns.

Inference Complete

Review metrics and download the trained prototype.

Total Accuracy

0%

Architecture

ResNet

Classes

0

Live Evaluation Hub

Drop image to test

Prediction Matrix

Sample Ground Truth Prediction Status