Tensorflow 1.5 Object Detection :: TFRecord


What is MobileNetSSDv2?

MobileNetSSDv2 (MobileNet Single Shot Detector) is an object detection model with 267 layers and 15 million parameters. It provides real-time inference under compute constraints in devices like smartphones. Once trained, MobileNetSSDv2 can be stored with 63 MB, making it an ideal model to use on smaller devices.

MobileNetSSDv2 Architecture

The MobileNetSSDv2 Model essentially is a 2-part model. The first part consists of the base MobileNetV2 network with a SSD layer that classifies the detected image. In essence, the MobileNet base network acts as a feature extractor for the SSD layer which will then classify the object of interest.
MobileNetSSDv2 Architecture
Image in Courtesy of Matthijs Hollemans

MobileNetV2 Results

MobileNet V2 outperforms MobileNet V1 with higher accuracies and lower latencies.
MobileNetSSDv2 Performance
Image in Courtesy of Google AI

Further Reading

Training a TensorFlow MobileNet Object Detection Model with a Custom Dataset: https://blog.roboflow.ai/training-a-tensorflow-object-detection-model-with-a-custom-dataset/