Neural Networks (general infos)

• evaluation, training, testing
• Weights for Layer -> trained (Numpy-Arrays)
• transfer learning: train already established models with new data (most of the time the output layers will be trained again)
• supervised learning: alogithm get a label for each train data (german: überwachtes Lernen)
• deep learning network types:

• base networks:

• provides high-level features for classification or detection
• classification, if you use an entirely connected layer at the end
• examples: VGG16, ResNet-101, Inception V2, Inception V3, Inception, ResNet, MobileNet

• detection networks:

• remove fully connected layer from base network and replace it with detection networks
• examples: SSD, Faster R-CNN, R-FCN