Real time Object Detection and Tracking using Tensorflow-lite
-
Developed and deployed a lightweight object detection system, using the MobileNet SSD model (Tensorflow framework and inference engine TFLite ).
-
Performed model hyper tuning and Optimization of the model using full integer quantization to achieve high speed with similar accuracy.
-
Deployed the model in C++ to achieve higher speed and efficiency on edge devices(CCTV cameras with NPU).
-
Deployed a system to perform activity recognition, and find specific objects, in CCTV footage using image embeddings