[CppIndiaCon 2023] Using C++ for your Artificial Intelligence Solution by Kushashwa Ravi Shrimali

Опубликовано: 29 Сентябрь 2023
на канале: CppIndia

403

It seems that AI is starting to pick up pace, but the fundamentals and challenges to get systems running for real-time purposes have remained the same. While Python has surely been the go-to language to train huge AI models, the ability of C++ to efficiently implement the protocols and perform the inference hasn't declined. From deploying models, communicating the data payload with cloud services, to efficiently merging with other pipelines in a product, C++ can do an amazing job. It's a fact, that the journey from having trained a model to detect accidents from a traffic feed to deploying the model in real-time in a traffic light system, is not so smooth. These times demand immediate results with decent accuracy, and hence performance is also a fair factor to judge a system's success. Using C++ for AI systems can help optimize the memory usage and performance. Consider a scenario of a traffic light detection based braking system. For a car driving at 30-40 km/h, a couple of additional seconds in model inference can pass or fail your entire ADAS function from the customer's perspective, irrespective of how good your model is. Storing and performing inference on the real time data is also a challenge, and how and when to leverage the available resources in a system is a good knowledge to have. This talk aims to give the audience a detailed overview on how to use C++ for their Artificial Intelligence systems or pipelines. Just being generally aware about the possible alternatives, goes a long way, and this talk will just do that.

Kushashwa Ravi Shrimali is a CSE Graduate (‘20) at IIIT NR, India (Gold Medalist, Transformational Leadership). He is currently working as a Backend Engineer at Abnormal Security. He is a lead contributor to Lightning Flash, and has contributed previously to PyTorch and OpenCV where he used C++ extensively for his contributions. He has also written several blogs on using PyTorch in C++ for deep learning. He has previously worked as a Research Engineer at Lightning AI, Software Developer at Quansight (developing PyTorch full-time with Meta), SDE (PyTorch Dev Team) intern at NVIDIA (HQ Santa Clara, US), AI Engineer at CareAI, and R&D Intern (Deep Learning) at ROSE Labs (NTU Singapore).

To join this community, please fill this form.
Registration form: https://cppindia.co.in/join_us
Webpage: https://cppindia.co.in
LinkedIn: / cppindia
Discord: / discord
Email: [email protected]