Enhancing the performance of AI models by 7X on the existing hardware using Edge Neural platform
Edge Neural worked with A-eye on the Object Detection Model. The model architecture is a proprietary Convolutional Neural Network built on Tensor Flow framework. The model identifies and locates the objects by drawing bounding boxes around these detected objects in a high-resolution video.
A-eye is an AI-based technology start-up that is on a mission to revolutionize multiple industries by making their solutions at least 4 times better using Computer Vision. It has built products focused on security, smart cities, traffic enforcement, OMR sheet verification that is powered by artificial intelligence.
It has built one of the highest accurate number plate recognition systems in the country. Their client base consists of hotels, hospitals, education departments, traffic police, tollgate, IT companies and IT parks.
- The task involves heavy computation process
- Require several GPUs for the scaled-up version in production environment
- Increase in cost
- Nvidia Jetson TX2 GPU
- Nvidia RTX2060 GPU
ENAP is the Edge Neural’s AI platform. It is a unified, cloud-neutral and hardware-agnostic platform. It integrates workflow to automatically build optimised AI models that are smaller and efficient. The platform provides enterprises, AI engineers and start-ups a tool to deploy DNN (Deep Neural Networks) on resource constraint, low- power and cost-effective hardware as well as manage the models without compromising on the performance.
Optimize the AI model for deployment on edge device which results in
- Lesser number of GPUs required
- Reduction of cost
The A-Eye uses many complex deep learning models for its products. This involves many steps, such as defining the problem, data engineering, algorithm decision model training, model evaluation and deployment to production. In the production environment, the model would be scaled up and would serve many customers simultaneously. So, A-eye had to use several GPUs which would cost an enormous amount. The more efficient and optimized the model is, the lesser would be the number of GPUs required thus reducing the costs.