What training methods were used to develop and train the AI system?
Aiconix uses metric learning, CNN fine-tuning, and few-shot learning with LLMs to train its AI system, while ensuring that customer data is never used for training through on-premises deployments.
-
Few-shot learning-based models (e.g., Face Recognition, Landmark Recognition, Speaker Identification, Logo Recognition): Training is based on a metric learning approach. For each training sample (e.g., an image), mathematical vectors (representing the concept) are extracted. These so-called embeddings are stored in a vector database for inference using k-nearest-neighbor search.
-
Fine-tuning of CNN layers for classification and regression tasks
-
Few-shot and context learning with large language models (LLMs)
-
Data privacy notice: Currently, we ensure with our partners that no customer data is shared or used for training. Only on-premises installation options are used for these scenarios.