arthurslima / w251hw7

Describe your solution in detail. What neural network did you use? What dataset was it trained on? What accuracy does it achieve? I used the tensorflow lib trained on the frozen_inference_graph_face dataset. The mean score on the benchmark was of 0.08

Does it achieve reasonable accuracy in your empirical tests? Would you use this solution to develop a robust, production-grade system? It performed well on the empirical tests. This solution is a bit less robust then using openCV since it does not have the same frames per second as the openCV

What framerate does this method achieve on the Jetson? Where is the bottleneck? We were able to run a prediction every 0.3s, making it perform 3 frames per second. The bottleneck is the neural net Which is a better quality detector: the OpenCV or yours? OpenCV is performing better than my own.