Researchers at Baidu have developed a lightweight CPU network based on the MKLDNN acceleration strategy, named PP-LCNet, which improves the performance of lightweight models on multiple tasks.
Inference speed on mobile devices based ARM architecture or CPU devices based x86 architecture has been challenging to get with the increase of model feature extraction capability and model parameters. Even many good mobile networks have been proposed to resolve this issue, but the speed of these proposed networks is not good enough on the Intel CPU due to various limitations of the MKLDNN.
Baidu’s new system improves the performance of models in many tasks, making it perfect for future artificial intelligence (AI) systems. The research group reinvented the lightweight model elements for a network designed on Intel-CPU.
PP-LCNet seems to outperform most of the state-of-the-art models, and for downstream tasks of computer vision, performs very well for tasks such as object detection, semantic segmentation, etc.
The research team focused on the following three questions to resolve:
- ‘How to promote the network to learn stronger feature presentations without increasing latency.’
- ‘What are the elements to improve the accuracy of lightweight models on CPU.’
- ‘How to effectively combine different strategies for designing lightweight models on CPU.’
The research states that its main contribution is summarizing a series of methods to improve the accuracy without increasing inference time and how to combine these methods to get a better balance of accuracy and speed.
The proposed network, PPLCNet, has a better speed balance and shows strong performance with better results on computer vision tasks. This research network also reduces the search space of NAS and offers quicker access to NAS with lightweight models. The research experiments in this research were implemented based on PaddlePaddle.
PPLCNet‘s practical image recognition system consists of detection, feature learning and retrieval modules, widely applicable to all types of image recognition tasks such as product recognition, vehicle recognition, logo recognition and animation character recognition.
Baidu claims that with this information, researchers can come up with several general rules for designing lightweight CNNs. It also provides new ideas on how other people can build their networks if they’re trying to get better models faster and find architectures that work best for them.