Project information

• Category: MLOps
• Client: Waletech (INSTAR Germany) Shenzhen, China
• Project date: 01 Feb, 2020
• Project URL: INSTAR Cloud

Replacing an old custom Tensorflow Embedding model that was used to prepare the numerical embeddings for an SVM classifier with a simple Principal Component Analysis (PCA) with a Histogram of Oriented Gradients (HoG) fallback. This task had more complexity as it required a custom training path to be triggered by customers through the Cloud UI. The used YOLO RTLite model provided the detected faces as crops that could be assigned to classes that then were used to train the pca-svm or hog-svm classifier pipeline.