Training Domain Specific Models for Energy-Efficient Object Detection
This work addresses energy-efficient object detection for specific applications like fixed-view intersections, but it is incremental as it builds on existing distillation and dataset reduction techniques.
The paper tackles the problem of achieving high accuracy and computational efficiency in object detection for limited domains by proposing domain-specific models (DSMs) trained with distillation and dataset culling. It results in DSMs surpassing COCO-trained models in accuracy for the same size and reducing training time by 93% with only a 3.6% accuracy drop.
We propose an end-to-end framework for training domain specific models (DSMs) to obtain both high accuracy and computational efficiency for object detection tasks. DSMs are trained with distillation \cite{hinton2015distilling} and focus on achieving high accuracy at a limited domain (e.g. fixed view of an intersection). We argue that DSMs can capture essential features well even with a small model size, enabling higher accuracy and efficiency than traditional techniques. In addition, we improve the training efficiency by reducing the dataset size by culling easy to classify images from the training set. For the limited domain, we observed that compact DSMs significantly surpass the accuracy of COCO trained models of the same size. By training on a compact dataset, we show that with an accuracy drop of only 3.6\%, the training time can be reduced by 93\%. The codes are uploaded in https://github.com/kentaroy47/training-domain-specific-models.