Resource-Efficient Object Detection by Sharing Backbone CNNs

Abstract

The detection of objects in image and video has made huge progress in recent years due to the use of deep convolutional neural networks (DNNs), with some network architectures becoming de-facto standards. This paper addresses the problem of sharing a backbone CNN for different tasks, for example, to enable detection of additional classes when an already trained network is available. When using multiple such neural networks, sharing a backbone can save inference time and memory consumption. We study sharing a common backbone between neural networks trained for different tasks (logoness and text block detection) based on Yolo v3. We provide results on the impact of different lengths of the shared backbone on performance and resource efficiency.

Cross-layer host-network co-design for QoS in streamed video data networking

“…However, when a lightweight model is derived from a compressed version of a complex model, specific training techniques designed for such compressed models, e.g., quantizationaware training [157], can be incorporated in the retraining phase to optimize continuous learning. Moreover, multiple models can share the same backbone to reduce memory consumption [9]. In our case, spatial aggregation in aggregated model training allows multiple video streams to have the same backbone parameters, so that only one copy of model backbone can be maintained in the GPU memory on each edge server, and memory consumption can be further reduced.…”